编程辅导 C C++ Java Python MIPS Processing 网络家教 在线辅导

   Implement a Simple Content-Based Image Retrieval System   Description:   CSS 490 B Multimedia Data Processing This project is to implment a simple Content-Based Image

 Implement a Simple Content-Based Image Retrieval System

Description:

  CSS 490 B Multimedia Data Processing

This project is to implment a simple Content-Based Image Retrieval system based on two different color histogram comparison methods.

1.     Test Image Database

This test image database includes 100 true-color images in .jpg format.

2.     Color Histogram

Color histogram comparison is a simple but effective apporach in CBIR systems. Here are two ways to combine the information from 3 color channels (R, G, B):

A.    Intensity Method

                                                                     (1)

By this way, the 24-bit of RGB (8 bits for each color channel) color intensities can be transformed into a single 8-bit value. The histogram bins selected for this case are listed below:

H₁:; H₂:;  H₃:;

H₄:      ……      ……      ……      H₆:;

… … …

H₂₅:;

B.    Color-Code Method

The 24-bit of RGB color intensities can be transformed into a 6-bit color code, composed from the most significant 2 bits of each of the three color components, as illustrated in the following figure.

       The 6-bit color code will provide 64 histogram bins.

For example, the R, G, and B values for a pixel are 128, 0, and 255 respectively. So the bit representations of them are 10000000, 00000000, and 11111111. Then the 6-bit color code value will be 100011.

In color code, there will be 64 bins with H1: 000000, H2: 000001, H3: 000010, … H64: 111111

3.     Histogram Comparison

You need to implement the distance metrics for histogram comparison. Let H_i(j) denote the number of pixels in j^th bin for the i^th image. Then the difference between the i^th image and k^th image can be given by the following distance metric:

¨      Manhattan Distance

                                                                           (2)

where M_i*N_i is the number of pixels in image i, and M_k*N_k is the number of pixels in image k.

Preliminary demo (in groups):

Show the graphic user interface is up running. It should allow users to browse the image database, select the query image, and to view the retrieved images. At this stage, it is not required that the retrieval algorithm is properly implemented so the retrieved images can be any images from the database.