Music Through Image Processing

· Create a 128x128 binary synthetic image with the following text: “THE RAIN IN SPAIN STAYS MAINLY IN THE PLAIN.”

· Create a 128x128 image of the letter “A” using the same font and font size as the image in step 1.

· Load both images and take their Fourier transform. Element-per element multiply the FT of A and the complex conjugate of the FT of the text image. Take the inverse FT and display its output.

· Test the robustness of this character recognition technique by editing in noise or distortion to your text image.

TEMPLATE MATCHING.

Below are the corresponding results for template matching exercise. The first image is the loaded synthetic image for the text: “THE RAIN IN SPAIN STAYS MAINLY IN THE PLAIN”. The second image is the loaded 128x128 image of the letter “A”. The third image is the same as the first image but added with a gaussian noise to represent the noisy image. The last image is the result for the test of robustness of this character recognition technique. By comparing the last two images, we can see that after the character recognition was applied, observable peaks are observed at the locations where the character “A” was supposed to be found. There are five letter A characters and they have all been recognized correctly.

1) SYNTHESIA

For the synthesia part of the activity, one of the test images was loaded and a look up table for the corresponding frequencies for each number was created. The supposed to be go to process of this part was to combine the image segmentation through hsv color space from the previous activity and use the template matching for the character recognition. This will be able to identify the number through the character recognition by creating same size images of the 10 single digts, telling what frequency is to be played and from the color segmentation identify the length of time this frequency is supposed to be played. Unfortunately, there was a problem regarding the segmentation of the colors red, blue and green in the images. Even though the range was properly selected the colors were still not successfully segmented.