Run detection to compare the logo template across fourteen scales. The maximum normalized score proposes a rectangle, not a guarantee.
Repair the menu, not the whole image.
Find the arrowed PES2021 mark with real template correlation, inspect the exact mask, then let local inpainting rebuild the covered texture.
Local image operation
Detect. Refine. Erase.
Editable rectangleimage pixels
Confidence--
Template scale--
Mask pixels0
Changed pixels0
Source checksum--
Output checksum--
Loading the local image engine.
What the repair knows.
Detection, selection, and reconstruction are separate decisions. A strong score can locate a logo while a careless mask still damages the menu.
The red overlay is the exact pixel set that may be reconstructed. Drag or enter coordinates when the automatic box is wrong.
Telea inpainting propagates nearby color and structure into the mask. Smaller, well-bounded regions usually preserve seams better.
Repair reasoning
Confidence is not quality.
Correlation answers “where does this template look similar?” The comparison blade answers the different question: “did the reconstructed texture actually preserve the menu?”
Low confidenceUse the manual rectangle; do not pretend detection succeeded.
Mask touches an edgeExpect weaker context because inpainting has neighbors on fewer sides.
Texture smearsTighten padding or reduce radius, then compare again.
