[BUG] Even when zoomed to 100%, the image is still compressed by the algorithm.

[signin404]

Describe the bug
Even when the zoom level is switched back to 100%, the image is still subject to algorithmic compression.
You have to manually drag the image to the bottom-left edge before you can actually switch to a 100% display without any algorithm-based interpolation.

To Reproduce
Steps to reproduce the behavior:

1. Open any image larger than the screen size.
2. Enable Lock window and Keep window size on navigation
3. Toggle zoom to non-100%.
4. Switch the zoom level to 100%.
5. At this point, even though the UI displays 100% zoom, the image quality is still compressed by the algorithm
6. When the image is moved to the bottom-left corner, the blurriness caused by the algorithm compression immediately disappears, allowing the image to truly switch to 100% display.

Expected behavior
Immediately stops any algorithmic interpolation when scaling is switched to 100%.

Screenshots

Environment Information (CRITICAL)

• QuickView Version: v6.29.0.0
• OS Version: Windows 10 21H2 19044.2006
• CPU: 13600KF
• GPU: 4070Ti
• RAM: 32GB
• Display: 1920x1080 100% Scale

Image Information

• Format: JPEG
• Resolution: 2609x1265
• File Size: 1.82MB

Additional context
QuickView.ini.txt
Recorded video: https://gofile.io/d/vURL4A
Test image: https://gofile.io/d/yULEdl

Can we add the feature of panning images with right-click drag?
Is it possible to disable the save dialog when switching to the next image after rotating an image or when exiting the program? I will never use it.

Why is there a constant CPU overhead even when the program is idle after rendering is complete?
Debugging the HUD: FPS still fluctuates even when idle.
Does this mean the program is still rendering and outputting new frames even when idle?
If so, can you optimize it so that it doesn't output new frames after rendering is complete to reduce the idle CPU overhead (Cycles delta) to the same 0 as MassiGra?

Would you be interested in adding the ability to interpolate using AI models?
For example, if the zoom level doesn’t change within 10 seconds, the displayed image will be replaced with one interpolated using AI.
When performing rapid zooming, traditional algorithms are still used to maintain real-time performance.
Model and dependency files can be large
QuickView can simply provide an interface for users to download the model and dependency files themselves if they need to.

[bananakid]

It looks like something reported previously in #221 (comment). However in my case dragging to the bottom left corner doesn't help to get a clear non-compressed image at 100% zoom.

Thanks for bringing up resource usage when idling issue! What software are you utilizing for tracking?

P.S. I think AI is great helper in day-today life, but adding this kind of feature at this time is unnecessary as QuickView isn't mature yet. Currently you can easily setup mpv as a lightweight image viewer with support of NN scaling models if you need it.

[signin404]

I use System Informer to track resource usage.

AI interpolation may be able to call NVDIA VSR directly to avoid downloading huge dependencies and model files?

[cmixed]

Although I personally use an NVIDIA GPU, QuickView should remain accessible to everyone—including AMD and Intel GPU users. Adding features that rely on vendor-specific tech (like calling NVIDIA VSR directly) would create a significant maintenance burden and lead to a fragmented experience across different hardware. Supporting all GPU vendors properly is already challenging enough without adding AI model deployment into the mix.

I think keeping QuickView focused on fast, reliable, and hardware-agnostic rendering with traditional methods is the safer bet for now.