Fundamentals of Computer Vision

Read an image, convert it to grayscale, copy it as image two, create a bright region of interest, and use absolute difference to obtain the subtraction result for region-of-interest processing.

Use image multiplication with masks to isolate the region of interest and assist segmentation and blending. Create a mask, perform element-wise multiplication with the original image, and extract the roi.

Explore image averaging in Matlab by reading a base image, adding Gaussian noise, and computing an average over several noisy samples to reduce noise.

Explore digital images, create black and white test images, and demonstrate logical operations: and, or, xor, not, to show how spots affect results.

Learn histogram equalization by reading an image, plotting and normalizing its histogram, applying equalization, and comparing the original and equalized images side by side, with adaptive histogram equalization introduced.

Demonstrate adaptive histogram equalization and its advantage over histogram equalization for image contrast in MATLAB. Compare grayscale results, histograms, and normalized histograms of adaptive versus histogram equalization.

Apply histogram equalization in Python with OpenCV to grayscale images, then implement contrast limited adaptive histogram equalization (CLAHE) and compare results using clip limit and eight by eight tiles.

Explore convolution with linear spatial filters, including the average and Gaussian filters, learn kernel rotation, zero padding, and full, same, and valid convolution methods.

Explore how average (box) filters blur images through convolution with a box kernel, compare to sharpening via local derivative, and assess padding and conv2 usage.

Compare the gaussian filter with the box filter, noting the weighted bell shaped kernel that preserves edges while smoothing, and recall the formula g(r)=k e^{-r^2/(2 sigma^2)} with Matlab implementation.

Apply a gaussian filter in Matlab using f spatial and m filter with a 21 by 21 kernel and sigma 3. Explore imfilter padding options to blur edges smoothly.

Explore Gaussian low-pass filtering with two kernels, observe high-pass, band-pass, and band-reject effects on a zone plate image, and compare results via montage in Matlab.

Learn how padding types—no, zero, full, mirror, replicate, circular, and symmetric—shape 2d convolution results, sizes, and border artifacts, and contrast correlation versus convolution in image processing and cnn.

Explore how circular, symmetric, and replicate padding affect convolution with a box kernel, comparing full and same padding and the resulting edge borders.

Explore nonlinear filters by applying the Laplacian kernel for high-boost sharpening and edge enhancement, using four connectivity or eight connectivity, implemented via conv2 on a sample image.

Apply unsharp masking and high boost filtering using a gaussian blur, form a mask from original and blurred images, and add it with a tunable k to sharpen edges.

Apply the median filter, a non-linear filter, to remove salt-and-pepper noise while preserving edges, demonstrated in Matlab with grayscale images and noise added via imnoise.

Examine the standard deviation filter, a non-linear tool that computes the neighborhood standard deviation (default 3×3, eight-connectivity) and how changing the neighborhood sharpens the image with Matlab's filter.

Learn how non-linear range and entropy filters operate on image neighborhoods, using Matlab to compute max-min ranges and entropy, compare outputs, and apply texture segmentation and edge/high-contrast detection.

Explore rgb, hsv, and lab color spaces, and learn how color channels, hue, saturation, and value drive color segmentation and practical matlab demonstrations.

Learn how to perform RGB color thresholding in MATLAB by extracting red, green, and blue channels, converting to double, and creating masks to segment specific colors such as red features.

Switch to the lab color space for color thresholding, extract the L, a, and b channels, build a red mask, then compare with rgb segmentation for clearer results.

Explore color thresholding in the HSV color space, extracting hue, saturation, and value to isolate red objects, tune thresholds, and reduce noise while comparing rgb and lab baselines.

Learn to threshold colors using Matlab's color threshold app, choosing hsv or lab color spaces, selecting orange with a polygon, and exporting a create_mask function for visualization.

Explore color channels in Python by examining jpeg's three channels and png's alpha channel, and learn to read, split, and inspect channel data with OpenCV.

Explore edge detection in Matlab using default and methods such as Sobel, Canny, and log on sample images, comparing results to understand how method choice affects detail and noise.

Explore edge detection techniques such as Sobel, canny, log, probit, and Roberts. Learn how changing threshold values and ranges improves edge visibility while reducing noise and discontinuities.

Explore morphological operations including erosion, dilation, opening, and closing on a synthetic image, using disk-based structuring elements to remove noise and reveal cracks through binary processing and overlay analysis.

Apply erosion, dilation, opening, and closing with various structuring elements (rectangular, square, diamond, disk, sphere) and sizes to illustrate morphological operations and compare results using Wiener filtering and hole infill.

Explore how to analyze binary images with regionprops in MATLAB, extracting area, perimeter, centroid, and bounding box of connected components using eight-connectivity and related properties.

Explore how bw area open removes connected components below a pixel area in binary images, using areas of connected components sorted in descending order to set a noise threshold.