4 ms·
I don't understand this statement. I recently received a similar reaction after talking about some of my research, so I'm wondering if this is a common approach
by waiquoo 11y ago
I don't understand this statement. I recently received a similar reaction after talking about some of my research, so I'm wondering if this is a common approach to new ideas in the startup world.
How is the algorithm bogus (assuming that there isn't a mistake somewhere and that the results are validated)? Just because the output in the examples is similar to other edge detectors doesn't mean the work is 'bogus' or worthless. It's a new algorithm, a new way of doing things. It may have advantages in specific areas over your 'trivial contrast-enhancing filters'. For example, there are many sorting algorithms, but you can get the same results regardless of the algorithm. Does this mean that any sorting algorithm other than the most trivial is 'bogus'?
- enriquto 11y agoSorry, I didn't mean to say "bogus algorithm". What I wanted to say is that the results look equivalent as those that can be obtained by linear filtering and rescaling the input image (a standard retinex-like technique): Compute minus the laplacian of the image, smooth the result using a cauchy kernel, linearly adjust the dynamic range. You can see the interior of the light bulb and the texture of the wall at the same time. I teach image processing, this is a standard exercice.
- dnautics 11y agoWell, in this case, enriquito's intuition is correct. The algorithm is very similar (if not identical[0]) to the algorithm found in the simplest edge-detection filter, which is to do isolate medium-band frequency contributions in phase space. https://github.com/JalaliLabUCLA/Image-feature-detection-using-Phase-Stretch-Transform/blob/master/PST.m https://github.com/JalaliLabUCLA/Image-feature-detection-usi... [0] I would be surprised if no one has ever used this exact arctangent kernel (with maybe different normalization and parametrization) in phase space.