Indian Institute of Information Technology, Allahabad

Image and Video Processing (IVP)

Aug-Dec 2022 Semester (MTech)

Course Information

Course Description: This course provides the basic understanding of the digital image formation and visualization, the visualization of relationships between spatial and frequency, the understanding of mapping the signal processing techniques to the digital image, an idea of multimedia data (image, video), and an exposure to various image and video compression standards.

Course Outline (Topics): The following list of topics is tentative. Based on available time slots, some topics may be dropped or added or reordered.

Unit 1: Digital Image Fundamentals- Simple image model, digital image formation, sampling, quantization, resolutions and representation, relationship among pixels, types of digital images. Color Image Processing: Color Representation, Chromaticity Diagram and Color Spaces, types of digital imaging and application areas. Enhancement- Point Processing: Contrast Stretching, Power-law and Gamma Transformation. Histogram Processing: Histogram Equalization and Matching.

Unit 2: Filtering and Restoration- Degradation function and Noise Models, Spatial Domain Filtering: Correlation and Convolution, Smoothing Linear and Nonlinear Filters: Mean and Median Filters, Adaptive Filtering, Sharpening Linear and Nonlinear Filters: Derivative, Laplacian, Unsharp Masking, High-boost Filtering. Frequency Domain Filtering: Filtering: Low-pass (Smoothing) and High-Pass (Sharpening), Ideal, Butterworth and Gaussian Filtering, Unsharp Masking and High-Boost Filtering, Homomorphic Filtering, Periodic Noise Reduction and Inverse Filtering and Wiener Filtering.

Unit 3: Edges, Lines and Boundary Detection- First and Second Order Edge Operators, Multi-scale Edge Detection, Canny Edge Detection Algorithm, Hough Transform: Line and Edge Detection, Morphological Operations and Application: Boundary, Skelton, Convex-Hull, Thinning, Pruning etc. Segmentation and Feature Extraction: Model-based and probabilistic methods and Image Classification Optimal and Multilevel Thresholding, Gray Image Segmentation, Watershed Algorithm.

Unit 4: Compression: Lossy and Lossless compression techniques, JPEG, JPEG2000 and Variants, Introduction to video processing, Compression standards and formats (MPEG and H.XXX), Video Streaming.

Course Instructor

Dr. Shiv Ram Dubey

TAs

  • Shivani Mohan Agarwal - MIT2021036
  • Akshay Jain - MIT2021038
  • Ayush Dubey - MIT2021040
  • Trinetra Devkatte - MIT2021096

Class Schedule
Class: Wednesday 10.00 AM - 12.00 Noon, Tute: Thursday 09.00 AM - 11.00 AM, Lab: Tuesday 03.00 PM - 05.00 PM
Course Ethics
  • Students are strictly advised to avoid the unethical practices in the course including review tests and practice components.
  • It is best to try to solve problems on your own, since problem solving is an important component of the course.
  • You are allowed to discuss class material, problems, and general solution strategies with your classmates. But, when it comes to formulating or writing solutions you must work/implement by yourself.
  • You are not allowed to take the codes from any source, including online, books, your classmate, etc. in the assignments and exams.
  • You may use free and publicly available sources (at idea level only), such as books, journal and conference publications, and web pages, as research material for your answers. (You will not lose marks for using external sources only at idea level.)
  • You may not use any paid service and you must clearly and explicitly cite all outside sources and materials that you made use of.
  • Students are not allowed to post the code/report/any other material of course project in public domain or share with any one else without written permission from course instructors.
  • We consider the use of uncited external sources as portraying someone else's work as your own, and as such it is a violation of the Institute's policies on academic honesty.
  • Instances will be dealt with harshly and typically result in a failing course grade.
  • Cheating cases will attract severe penalties.

Schedule

Date Topic Class Material
L01-08: Aug 24-25, 2022, Aug 31, 2022, Sept 01, 2022 Introduction and Fundamentals Slide
L09-10: Sept 07, 2022 Intensity Transformation Slide
L11-12: Sept 08, 2022 Filtering in Spatial Domain Slide
L13-16: Sept 14-15, 2022 Filtering in Frequency Domain Slide
L17-18: Sept 21, 2022 Image Restoration Slide
L19-22: Oct 12-13, 2022 Morphological Image Processing Slide
L23-25: Oct 19, Nov 02, 2022 Image Segmentation Slide
L26: Nov 02, 2022 Color Image Processing Slide
L27-30: Nov 09-10, 2022 Image Compression Slide
L31: Nov 16, 2022 Image Compression Standards Slide
L32: Nov 16, 2022 Video Processing Slide

Computational Projects Added to Teaching Laboratories

Project ID Team Project Title Abstract
IVP22_P01 Raj Jaiswal (MSG2022032), Arvind Kumar (MSG2022033), Vikash Rajput (MSG2022035) Detection of Fake Currency using Image Processing As we all know, a country’s ”currency” is very important. All the economic activity of that country is dependent on the currency itself and how it is circulated. With the advancement of printing and computer vision technology, much fake currency has been circulating in the country. In our project paper, we examine Indian currency to determine whether it is genuine or counterfeit. Many illegal activities have been carried out in a country like India using counterfeit currency, and counterfeit currency is also used by terrorists for funding and other purposes, which is a serious issue in a country like India. Our government has implemented several strategies to reduce the use of counterfeit currency, but the problem persists. In recent years, we all know the government has also changed the currency of India to minimise the use of fake currency, but the bad people always find a way to circulate the false currency. In our project, we have used several processing techniques for image and video processing (IVP) to detect whether a currency is fake or not. In its recent report, the RBI said on March 31, 2022, that out of the total number of fake Indian currency notes detected in the banking sector, 6.9 percent were detected at the RBI and 93.1 percent at other banks. From this data, you can understand how much it is necessary for the currency to be detected, whether it is fake or not.
IVP22_P02 Kavathiya Khyati H (MSG2022017), Harsh Kag (MSG2022028), Ankit Raj Ravi (MSG2022030) Fake Currency Detection using Image Processing For people and corporations, Fake currency detection is a serious issue. Color printing technology has sped up the printing and mass manufacture of counterfeit banknotes. Anyone can learn to produce cash notes with the highest level of precision using a basic laser printer. False notes being used is one of the major issue to address. Tragically, India has struggled with concerns like corruption and black money which is also concerned with fake currency issues. Technology advancements have increased the chance for bogus money to be fabricated and spread, which has a harmful effect on the economy overall. This problem leads to the developing technology which can be easily used to detect fake currency notes. Equipment is available in banks and other commercial places to confirm the legitimacy of the money. There is a need for software that can be used by common people to detect fake currency because the typical individual does not have access to these systems. The suggested system employs image processing to identify genuine from counterfeit cash. Python programming is used exclusively in the system’s design.
IVP22_P03 Samar Jyoti Das (MSG2022020), Hritu Raj (MSG2022018), Bhargav Burman (MSG2022034) An improvement on single-image based DCP We propose an improvement on single image enhancement for de-hazing by using dark channel prior or DCP. Dark channel prior has some shortcomings. One of them is that dark channel prior and other methods based on it fail to process brightness hotspots. In images containing large areas of high brightness, most commonly the sky area, the algorithm will fail to remove hazing and keep the characteristics of the original image intact. So method to section the sky and ground area has been considered. This will enable DCP to handle images with prominent sky areas and allow DCP to be viable to more real world situations and applications.
IVP22_P04 Harsh (MSG2022024), Harshit Gupta (MSG2022026), Umesh Maurya (MSG2022029) Image Deraining using Image Processing In this work, we propose a method for single image deraining. Unlike movies, eliminating rain from a single photograph is a difficult challenge since there is no temporal information provided. The suggested image deraining approach here consists of two phases. In the first step, we use a guided filter or bilateral filter to obtain a coarse rain-free image. This serves as a guided image for the second stage, in which we employ a L0 smoothing filter inspired by L0 gradient minimisation to generate the final output image.
IVP22_P05 Diya Srivastava (MSG2022004), Pragati (MSG2022013), Ashutosh Verma (MSG2022019) Image Inpainting using Image Processing Images Inpainting is an important topic of research in the field of image processing. The prime goal of image inpainting is to recover missing details in an image, demosaicing etc. Image inpainting holds broad applications. The wide application of image inpainting in the field of forensics, criminal detection etc has made image inpainting a subject of contineous interest and research. The recent, COVID-19 pandemic has provided a new set of opportunities in the image inpainting domain. The detection of face under the face mask seems like a future endevour for the domain. In this paper we have discussed about out project on the image inpainting which has been worked under the environment of python using the inbuilt python libraries.
IVP22_P06 Hrishav Raj (MSG2022002), Amit Roy (MSG2022023), Sumit Bhimte (MSG2022025) Low Light Image Enhancement The compressed dynamic range of low-light images can be expanded to provide fine detail information. A challenging task in the field of image processing is enhancing the contrast of a low-light image. The methodology for improving low-light images is suggested in this paper. While preserving the original image’s details, this technique can enhance the contrast and brightness of low-light images. Maintaining image brightness and keeping image details can help with image enhancement. Histogram equalization can be used to improve various low-light image types. In order to enhance images without losing any of their details, The impact of conventional histogram equalization is controlled using the Dynamic Histogram Equalization approach. Traditional histogram equalization techniques may result in extreme enhancement and detail loss, creating an unclear and artificial image. Changing Histogram Prior to equalizing each partition separately, the equalization process divides the histogram of image on the basis of local minima and mark out distinct grey level ranges to each portion. To further verify is there any dominating portions are present or not, these partitions are put through a repetition test. This technique outperforms other current methods by effectively enhancing contrast without causing serious side effects.
IVP22_P07 Vivek Kumar Soni (MSG2022009), Shyam Dongre (MSG2022016), Akash Tyagi (MSG2022027) Image Dehazing using Image Processing The restoration of images that have been distorted by various degradations is one of the major issues in the field of image processing. Images of outdoor images taken in poor weather often have atmospheric deterioration like haze, fog, and smoke because the atmospheric medium’s particles absorb and scatter light as it moves from the scene point to the observer.In poor weather conditions, the colour and contrast of the acquired image degrade as a result of the presence of these air particles. This could make it challenging to identify items in the hazy sceneries or photographs that were shot. The ability to enhance and remove haze from outdoor photos is now available because to recent advancements in the field of computer vision. In order to recover better and improved quality haze free photographs, this paper presents various haze removal procedures that can be used to remove the haze from the obtained hazy images.
IVP22_P08 Harshit Gupta (MSG2022038), Naveen Sharma (MSG2022039), Dasaroju Jagannadhachari (MSG2022006) Image Search from Hindi, Spanish, and French Text Data The aim of our project to automate the application to overcome from the language barrier among different images. This paper is intended to explore the Image Search from Hindi, Spanish, and French Text Data. Image search help us to extract Hindi, Spanish and French texts from the images and scanned documents so that it can be edited, formatted, indexed, searched, or translated.
IVP22_P09 Bhavesh Kumar Bohra (MSG2022031), Aditya (MSG2022036), Manish Kumar (MSG2022037) Road Lane Line Detection using Image Processing This project is intended to build a machine learning project to detect lane lines in real-time. We will do this using the concepts of computer vision using OpenCV library.It involves real-time image and video processing concepts along with ML algorithms. To detect the lane we have to detect the white markings on both sides on the lane.Detecting lane lines is a fundamental task for autonomous vehicles while driving on the road. It is the building block to other path planning and control actions like breaking and steering.
IVP22_P10 Mohd Faiz Ansari (MSG2022011), Rakshit Sandilya (MSG2022014), Aditya Vaishy (MSG2022012) Satellite Image Analysis using Image Processing Satellite images are one of the most important tools to keep track of various human activities and assets of the earth like vegetation, water bodies, air, climate change, etc. There are many artificial satellites revolving around the orbit of our planet which is the source of images. These images are exposed to noise and irregular illumination which are undone by image processing techniques. This paper talks about different techniques that can be used to enhance a satellite image to fetch more details from it and its analysis can be used for a better cause in the future.
IVP22_P11 Sayantan Chakraborty (MSG2022001), Dipankar Karmakar (MSG2022003) and Aditya Ramesh Patil (MSG2022007) Disguised Face Recognition Using Image Processing Over the past two decades, face recognition research has grown astronomically. The current face recognition systems attain extremely high accuracy on large-scale unconstrained face datasets by starting from algorithms capable of conducting recognition in confined contexts. The majority of face recognition systems are vulnerable to failure under disguise fluctuations, one of the most difficult covariates in face identification, even though new algorithms continue to gain performance improvements. On existing disguise datasets, certain algorithms have been shown to produce promising results in the literature; however, the majority of these datasets are made up of photographs with little alterations that were frequently taken in controlled environments. The Disguised Faces in the Wild (DFW) dataset, which contains over 11,000 photos of 1,000 identities with variations across various types of disguises, is a novel dataset that is proposed in this study.
IVP22_P12 Rupesh G (MSG2022022), Raj Ahamed Shaik (MSG2022021), Himanshu Mishra (MSG2022015) Restoration of Old Images by Removing Scratches In this work, we propose a method to detect scratches and removal from the images. The proposed method, combines bandpass, homomorphic filtering, Hough transform and image dependent thresholding. This allows us to reliably distinguish defects from intrinsic image features and achieve quality image restoration. The method is efficient and can be used in real-time applications.
IVP22_P13 Nikhil Rajput (MSG2022005), Himanshu Mittal (MSG2022008), Jyothi Krishna Behera (MSG2022010) Single Image De-raining using Image Processing The Non-Negative-Factorization approach is used for the pre-processing step of single picture de-raining. According to the claim, NMF can factorise a matrix A with dimensions m x n where each element is ≥ 0, into two matrices W and H. W and H contain matrices with non-negative entries and m x n and k x n, respectively. Before eliminating the rain, the recommended method separates the picture into high and low frequencies. Because NMF has excellent noise-filtering abilities, it processes the low-frequency component. However, the highfrequency component passes via Canny edge detection, dictionary filtering, and block copy to eliminate rain and retain the edge information.
IVP22_P14 Ajay Kumar Yadav (RSI2022502) Super-Resolution of an image with Deep Learning With Deep Convolutional neural networks, we suggest a Single Image Super-Resolution (SISR) model that is both very effective and quick (Deep CNN). Deep CNN recently demonstrated that they perform significantly better during reconstruction of singleimage super-resolution. Deeper CNN layers are becoming more popular as a way to boost performance. Deep models, on the other hand, need more processing power and are incompatible with network edge devices like smartphones, tablets, and Internet of Things sensors. Our model uses Deep CNN with Residual Net, Skip Connection, and Network in Network to deliver cutting-edge reconstruction performance with at least ten times reduced computation costs (DCSCN). Image features in both local and global regions are extracted using a feature extractor that combines Deep CNNs and Skip connection layers. Image reconstruction also uses parallelized 1*1 CNNs, such as the Network in Network CNN. This structure allows for quicker calculation with less information loss and the direct processing of original pictures by reducing the size of the output from the preceding layer. Additionally, we maximise the number of layers and filters in each CNN to greatly lower the cost of computation. As a result, the suggested method not only produces state-of-the-art results but also performs computing in a quicker and more effective manner.
IVP22_P15 Debesh Kumar Shandilya (RSI2022504) Removing Rain From an Image Rain is one of a common weather condition. It has severe adverse effects not only on human visual perception but also on the performance of different computer vision tasks. Mostly rain streaks and background of an image are closely related to each other. So to remove rain streaks from an image and finding background image while maintaining the accuracy of background image is a challenging task. In this research work my aim is to develop a model using Image processing tasks and Deep learning architecture to achieve the task.
IVP22_P16 Manish Rajput (RWI2022505) Image De-noising using Image Processing Images undergo various processes like acquisition, compression and / or transmission etc., and thus they are exposed for changes like getting infected with noise. This image contamination due to various factors can lead to loss of information and subsequently distortion of image. A distorted image when processed for analysis doesn't give a good result thus making Image denoising an important and vital step in mordent digital image processing. The process of removal of noise from Images is called Image denoising. The objective of the image denoising is to recover original image from the noise infected image. Since noise is a high frequency component hence it throws challenges during the denoising process. With the advent of modern techniques, this classical image denoising problem is now dealt with more accurately and precisely, leading to image restoration as closely as possible. There exists various techniques to eliminate image noise. In this project we will be focusing on the Convolutional neural network (CNN) methods to denoise an image.
IVP22_P17 Shikha Verma (RWI2022004) Image Processing of Satellite Imagery In real time, satellite images give a real view of the Earth and its ecosystem. They have been offering a great deal of information to scientists, meteorologists, and decision makers to help them make better policies and decisions.Image processing of satellite products is performed in order to enhance the image and to extract useful information. Satellite imagery along with image processing provides an assistance for Geographic Information System (GIS). Moreover, it provides the services for real-time applications such as monitoring weather, agriculture, change detection, resource identification, natural disaster prevention and so forth.

Grading

  • C1 (30%)
  • C2 (30%)
  • C3 (40%)

Prerequisites

  • Computer Programming
  • Data Structures
  • Problem Solving
  • Ability to deal with abstract mathematical concepts

Books/References

  • Digital Image Processing by Willam K. Pratt, John Willey & Sons
  • Digital Image Processing by Gonzalez, Rafael C., and Richard E. Woods, Pearson Education
  • The Essential Guide to Video Processing by Alan C. Bovik, Academic Press

Disclaimer

The content (text, image, and graphics) used in the slides are adopted from many sources for Academic purposes. Broadly, the sources have been given due credit appropriately. However, there is a chance of missing out some original primary sources. The authors of this material do not claim any copyright of such material.