I thought the new percept generation code would balance out the softness and structure / edges but it still looks about the same as before, only with a little more texture:
I’m wondering how I could combine the soft and stable with the bright and chaotic aesthetics. This is already the case to a degree, except most percepts contain many regions, and therefore the majority are quite soft. I can either render the existing clusters differently, or I can increase the number of clusters.
The current plan is to map from the area of each constituent region to the weight in the percept. Large regions would be less weighted than small regions, which would have sharper edges. Hopefully this would result in both more colour diversity (due to highly weighted clusters) and maintain stability (as those most unstable clusters are smallest). The rationale is that larger regions imply a lack of colour diversity in the input source, and areas covered by large regions are therefore less likely to contain salient information. Small regions, on the other hand, are likely to occur in areas of high salience. This approach should preserve some of visual aspects of high salient areas.
On another note, I got a strike for using copyrighted material in my last youtube post. I can only presume that my use of the original sound-track to aid in interpretation was the culprit. I hope to get this resolved by removing the audio from the video.