1.1. Use the example given in (bwdist) function to compare the 2-D distance transforms for each of the supported distance methods.
Each figure depicts the binary of the image along with its Euclidean, Cityblock, and Chessboard distance transforms. This comparison depicts how the different metrics affect the distance contours of each image. 1.2. Construct the distance transform for the image ‘leaf’ for each of distance methods. What aboutthe robustness to noise using ‘leafnoisy’ image? • Euclidean Distance Transform: The distance contours in the noisy image appear slightly distorted near the leaf's boundary due to the added noise, but the central structure remains relatively unaffected. • Cityblock (Manhattan) Distance Transform: The noise slightly alters the boundary's distance gradient, but overall, the transform remains stable, showing similar contours to the clean image. • Chessboard Distance Transform: Chessboard distance transformation also presents slight boundary distortions under noise, while keeping its main structure stable in the central regions. 2. Arithmetic operations
2.1. Implement a simple code to find the differences between two images. See the operator –between images, without any loop. Use it with images ‘original’ and ‘original2’. Comments withthe visual inspection... How then to count these five “errors”? It is not necessary to precisely do it,just assume the eventual problems and how to implement it. Estimation Approach: Thresholding: The use of a threshold will enable us to filter out minor, insignificant differences and tend to those that are larger. Counting: An estimate of the count of the errors is reached by counting the number of pixels above threshold. 2.2. Make now the difference between ‘chroOp’ and ‘chroL’. According to you, what kind of information is in that way highlighted? Try to mask the original image ‘chro’ with these results in color in order to highlight particular parts of objects directly. The image ‘result’ is the good one…cat function can be used if necessary to create 3 planes image from one with only 1. The following can be obtained from the result: Difference Image: This image points to the regions where chroOp and chroL differ. Contours around every highlight show where these changes take place, which might indicate the boundary of possible objects or structural changes.
Original Image with Highlighted Differences: Overlapping the edge-detected
differences in red onto the original chro image allows direct visualization and highlighting of parts of objects. The red outlines now emphasize the boundaries where chroOp and chroL differ, and this could be a possible area of interest or a region of structural changes. 3.Convolution
Define a signal f and convolve it with itself to get g. Repeat this
procedure three times and plot your results. Now, define a Gaussian convolution kernel (look at function fspecial)
Successive convolutions of the signal f with itself yield peaks that are progressively smoother and wider, each such convolution spreads and smooths the original signal. At each iteration this is increasingly a Gaussian-like shape typical of repeated self-convolution of a Gaussian shape. Convolve an image with the defined convolution kernel (look at function conv2). Test what you will obtain if you convolve the image with itself. Justify your observations. Original Image Section: Provides the original section image with fine pixelated detail. Gaussian-convolved section: The Gaussian convolution has smoothed out the details. It contains much less noise, but larger-scale patterns are enhanced. This is a general result of Gaussian convolution, where the high-frequency details become smeared out, while the general structure is preserved. 4.Dirac Delta Function Integrals 1. Integral 1: ∫₋∞⁺∞ δ(x)f(x) dx = 1.00
