Image processing.

Objectives: The objective of this assignment is to gain familiarity with basic image processing methods. It also provides an introduction to the common image processing and analysis tasks using OpenCV. After completing this assignment, you will learn how to: Open and read image files; Perform simple mathematical operations on images; Carry out translation based alignment; Perform image adjustment and restoration; Task: Super-resolution techniques enable us to create images with higher spatial resolution and lesser noise than regular camera images. They attempt to reconstruct the original scene image at high resolution (HR), given a set of observed images (might be only one) at lower revolution (LR). The need for high resolution is common in computer vision applications for better performance in scene analysis. The goal of this assignment is to write a program that takes a set of (at least three) input images and combines them into a single higher resolution image. Below is a set of three LR images taken of the same scene, the combined HR image with blur and noise, and the final result obtained after applying a super-resolution process. It is recommended that you use the OpenCV library either in C++ or Python . Instructions: Task 1 Photograph several normal or low resolution images of the same scene. You should then align the images to ensure correct appearance of the combined high resolution image. A simple way to perform the alignment is by searching through all possible offsets in some suitable range (e.g. 20 pixels for low resolution images) and computing for each a score measuring the quality of the match. Three suitable metrics include sum of squared differences (SSD), sum of absolute differences (SAD) and the normalised cross correlation (NCC), all of which are implemented in OpenCV. For this assignment it is sufficient to use an x, y translation transform but feel free to implement other methods. Task 2 Once you can align several low resolution images, the next objective is to construct a single high resolution image. To implement this, you will firstly put the registered LR images on a high resolution image grid, where non uniform interpolation methods are used to fill in those missing pixels on the HR image grid. Make sure you can display some meaningful intermediate results. Task 3 Having obtained the HR image, you will perform image deblurring by any classical deconvolution algorithm with noise removal to achieve the final HR image. For example, you can use a gaussian smoothing filter and subtract the smoothed version from the original image to get a sharp image. Or, you can apply bilateral filter on the image to reduce unwanted noise and keep edges sharp. These filters are implemented in OpenCV. Optional task: Try to improve the visual quality of the results of the basic algorithm, e.g. through the use of frequency techniques or Bayesian super-resolution approaches. Evaluation: Several low resolution image sequences (of three images) will be released on the day the assignment is marked. You will take your marker through the steps showing the output of the image alignment procedure, a representation of the high dynamic range image and the final outputi, as well as the results you have obtained on your own data. Tasks 1, 2 and 3 are sufficient to complete the assignment and will be marked against the maximum mark achievable. The optional task, if completed, will attract a bonus mark. Deliverables: In addition to demonstrating your work, you will also submit a two page report to your marker, in hard copy on the day. This report should explain briefly the approach you have taken in Tasks 1, 2 and 3 and include some sample input sequences and the results obtained. Optionally, include details of any enhancements you have implemented. Software: Download OpenCV and read guided tutorial : CourseMenu> CourseWork> Tutorials> OpenCV Intro and more detailed tutorials: [login to view URL]