4. Optimization - radhikahorti/Embedding-Synthetic-Pattern GitHub Wiki

This chapter discusses the techniques used for optimization of the selected solution.

4.1 Method of Optimization

The optimized solution for the given problem statement used is as follows: • Embedding the synthetic pattern onto the best matched block in terms of contrast Go through the image iteratively in terms of blocks and compare the contrast of each block with the synthetic pattern. Embed the pattern onto the best suited block. • Matching the contrast of the selected block by adjusting the contrast of the synthetic pattern Adjust the contrast of the pattern iteratively until it aligns within a threshold (in this case, within 5 units).Once the contrast is adjusted, embed the modified pattern into the specified block in the image. • Matching the mean intensity of the selected block by adjusting the mean intensity of the synthetic pattern Compute the intensity difference between the block and pattern, then normalize the pattern array. The function then adjusts the intensity of the pattern by applying the intensity difference and normalization.

4.2 Selection and justification of optimization method

A comparative study of all the above methods was carried out. The most above appropriate method to embed synthetic patterns onto the input image such that all the natural statistics are preserved is by embedding the pattern onto the block that has the same or similar contrast to it, as in this case the SSIM was found to 1.0 that is the image is virtually unchanged even though the pattern has been embedded onto it and the local statistics measurement block did not find any dissimilar blocks when both the images where compared. However, in case of practical scenarios it would be more appropriate to match the contrast/ intensity of the selected block to that of the synthetic pattern, as in these cases the SSIM score was near 0.99 which shows that the modified image is almost similar to the input image.