IT – 472 Digital Image Processing
Practical details Lectures – CEP 205 Lab – Lab 205 Grades: Monday 09:30 – 10:30 Wednesday, Thursday 08:30 – 09:30 Lab – Lab 205 Wednesday 14:00 – 16:00 Grades: Based on 1 or 2 internals, finals, assignments, labs, paper reading etc. Can change!
References & Prerequisites Digital Image Processing, by Gonzalez and Woods Fundamentals of Digital Image Processing, by Anil Jain Prerequisites Linear algebra Signals and systems: 1D Fourier transforms, convolution, sampling theorem.
Digital Image Processing DIP: Processing multi-dimensional signals. What all processes? Enhancement : Makes the signal more conducive for a specific task. Noise removal, Contrast stretching, Change brightness, Sharpening Restoration: Tries to undo a degradation process. Some reasons for degradation: Camera impulse response is not an impulse Relative motion between object and camera
Digital Image Processing Compression: Always easier to handle smaller data. Lossy and Lossless compression. Segmentation: To separate object of interest from ‘background’ Morphological Processes: Nonlinear Processing based on set theoretic concepts. Filtering, computing region descriptors
What after DIP DIP allows you to explore: Subjects offered at DA-IICT: Computer Vision Robotics Pattern Recognition Understanding Human Visual System Subjects offered at DA-IICT: Computer Vision (Autumn) Numerical Differential Geometry in CV (Winter) Pattern Recognition (Autumn) Image Analysis (Winter, under construction!)
Digital Images Image – 2 dimensional function f(x,y) Digital image – Sampled and Quantized image, represented by a matrix I(x,y) of size, say m x n. Each element is called a Pixel. Grey level digital image – the values of I(x,y) are discrete, usually from 0 to 255, 0 representing black, 255 representing white.
Digital Image Acquisition
Image formation model The imaging system senses amount of energy reflected/allowed to pass through by the object. 0 < f(x,y) < ∞ The energy reflected by the object comes from an illumination source. If i(x,y) is the energy incident at point (x,y) of the object and r(x,y) is the reflectivity of that point, then f(x,y) = i(x,y)r(x,y)
Sampling and Quantization Digitizing the coordinate value – sampling. Digitizing the amplitude – quantization. For an image of size m x n, with L=2k different grey levels, it requires m x n x k bits of storage: m = 1024, n = 1024, k = 8 (L = 256) 1 Mbyte At 1 Mbps, 8 secs to get 1 image!
Resolution Spatial resolution – smallest distinguishable detail in the image. Higher sampling Higher spatial resolution Grey level resolution – smallest distinguishable change in grey level