Shallow depth of field image dataset for image sharpness and fitting results of BPBD and other algorithms

Citation Author(s):
Xidian University
Submitted by:
zhixiang duan
Last updated:
Mon, 08/16/2021 - 03:19
0 ratings - Please login to submit your rating.


We provide two folders: 

(1)The shallow depth of field image data set folder consists of 27 folders from 1 to 27. 

In folder 1-27, each folder contains two test images and two word files. Img1 is the shallow depth of field image with the best focusing state taken with a 300 mm long focal lens, and img2 is the overall blurred image. 

The example to the weakness of the state-of-the-art methods word file shows the evaluation results and failure of various algorithms, and the subjective scoring results word file shows the results of the subjective evaluation of 30 volunteers.

(The subjective test process, in which evaluation attribute is set to sharpness, strictly adheres to the Performance Methods in Stimulus-comparison Methods (SC) in ITU-R BT.500-14, 

and the observer’s task is to judge which one is the sharper from two pictures in each group. The sharper picture is marked as 1 and the less one is marked as 0.)

(2)The fitting results of various algorithms(include our proposed metric BPBD) folder contains six folders showing the fitting results on TID2008, TID2013, LIVE, CID2013, FocusPath and CSIQ.


Readme contains a detailed description of the database and experimental results