The Sparsity and optimal sparsity threshold of images in the USC-SIPI dataset

Citation Author(s):
Wenhao
Liu
Submitted by:
Kehui Sun
Last updated:
Mon, 04/08/2024 - 22:00
DOI:
10.21227/nfp9-ah31
License:
25 Views
Categories:
Keywords:
0
0 ratings - Please login to submit your rating.

Abstract 

This dataset is used to compressive sensing scenes and assist users in training the optimal sparsity threshold. The input features are image compression ratio, image size, and sparsity, while the label field is the optimal sparsity threshold. The test images are from the USC-SIPI dataset. 

By performing DWT on image P of size N×N, one can obtain the sparse matrix A. Increasing the sparsity of matrix A appropriately can improve the compression performance in compressive sensing (CS) processing. The sparsification method is described as A(abs(A) < Ts)=0. Thus, the key issue is how to choose an appropriate threshold Ts.

Instructions: 

Steps:

1.      A = DWT(P)

2.      sparsity = sum(abs(A(:)) < Ts)/N^2

3.      A(abs(A) < Ts)=0

4.      CS sampling for matrix A

5.      OMP reconstruction for compressed image

6.      Evaluation of reconstructed image quality: PSNR

 

Feature:

CR: compression ratio

N: image size

Sparsity: sparsity of matrix A with different Ts.

 

Label:

Ts_opt: Optimal sparse threshold Ts that maximizes the PSNR value of the reconstructed image

Comments

update

Submitted by Kehui Sun on Mon, 04/08/2024 - 22:01