Matlab---Traffic-cone-detection-and-Color-Image-Processing

This report analyzes the fundamentals of color image processing, beginning with the exploration of transforming colors to different spaces, and extending this to segment and detect specific objects in an image. The object under study for this report is orange traffic cones. In the first section of the report, we study two different methods of detecting the cones in the image based on their color. The procedure adapted in the first method is to manually identify samples of the orange color cones in the images and compute the mean and standard deviation of the orange color in these images. We have compared two color spaces, the RGB and YCbCr space and observed that the YCbCr identifies objects better since we remove the illumination component from the image. The second method deals with formulating the conditional probability distribution of the pixels that represent and do not represent a traffic cone. We then test both the models by testing how accurate they are in detecting the cones in a set of test images. Comparison of the two methods indicates that probability distribution more accurately identifies the traffic cones. We also study the significance of histograms by comparing the histograms of the three color channels for different regions in an image and finding how similar or closer they are using the chi-squared distance measure. In the last section, we also analyze two image compression formats – JPEG and PNG, compare their compression techniques and analyze their quality. MATLAB has been used as the programming language.