Test Image Generation using Segmental Symbolic Evaluation

Abstract

Image processing applications have played a vital role in modern life and they are required to be well tested due to their significance and human dependence on them. Testing of image processing application is difficult due to complex nature of images in terms of their generation and evaluation. The presented technique is first of its type to generate test images based on symbolic evaluation of program under test. The idea is based on the fact that, neighboring image operations are applied by selecting a segment of image pixels called a window, and iterated by sliding window over entire image. We imitate neighboring operations using symbolic values for the pixels rather than concrete values. The path constraint is extracted for each path in the program under test and solved for concrete solutions. Test images are generated based on solution of path constraints for each identified path. We have tested the proposed scheme on different programs and the results show that test images are successfully generated for each path to ensure the path coverage of the program under test and identifying infeasible paths.

Adaptive Omni-directional Walking Method with Fuzzy Interpolation for Biped Robots

“…In rotation walking adjustment part, the relative acute angle between a robot and target is regarded as the input and the rotational speed is the output. Especially, the relative distance and acute angle can be obtained by some methods, such as image processing methods [30][31][32][33], localization techniques [34][35][36], mapping techniques [37][38], noise reducing method [39][40] and so on.…”

“…according to (33) with the r ε being the range of maximal allowable fuzzy distance. After the fuzzy output vector } ,..., 2 ,…”

Metamorphic Relations Based Test Oracles for Image Processing Applications

“…In the article by (Hui & Huang, 2013), the authors say that random testing is considered unbiased and empirically it is shown that metamorphic testing shows better results with random tests in terms of mutation score and fault detection ratio. In the article by (Jameel & Lin, 2014), the authors say that test cases with higher code coverage are considered more effective in terms of their error detection capabilities. We have selected successful test cases randomly in our experimentation.…”