Reducing blocking?

[Mr-Z-2697]

During some simple tests, I noticed that higher value of block_step will make more blocking when sigma is high, and block_step=8 will cause a few visible blocking even when sigma is relatively low (~5).

As mawen mentioned in this issue: HomeOfVapourSynthEvolution/VapourSynth-BM3D#18 (comment), using smaller block_step in final estimate will decrease some artifacts (including blockiness).

Is it possible to do in this implementation?

[Mr-Z-2697]

Misoperation, sorry.

[WolframRhodium]

Yes, it is possible by specifying block_step to a low value if blocking is observed. The parameter is presented in all of the implementations.

[Mr-Z-2697]

Yes, it is possible by specifying block_step to a low value if blocking is observed. The parameter is presented in all of the implementations.

Yeah, I'm already doing that.

But what I mean is "using block_step smaller than basic estimate in final estimate".

I mean if the "non-blockiness" of block_step basic and final (5,4) and (3,3) are close, that could still be some performance improvement, right?

[WolframRhodium]

Each bm3dcpu.BM3D function call only corresponds to one basic estimate or one final estimate. You may use it like

basic = bm3dcpu.BM3D(src, block_step=5)
final = bm3dcpu.BM3D(src, ref=basic, block_step=4)

for your purpose.

[Mr-Z-2697]

Each bm3dcpu.BM3D function call only corresponds to one basic estimate or one final estimate. You may use it like

basic = bm3dcpu.BM3D(src, block_step=5)
final = bm3dcpu.BM3D(src, ref=basic, block_step=4)

for your purpose.

Well, thank you so much!

I didn't know each function call olny corresponds to one estimate before.