Super-resolution from a single image
read more
Citations
Non-local Neural Networks
Photo-Realistic Single Image Super-Resolution Using a Generative Adversarial Network
Perceptual Losses for Real-Time Style Transfer and Super-Resolution
Image Super-Resolution Using Deep Convolutional Networks
Perceptual Losses for Real-Time Style Transfer and Super-Resolution
References
A Review of Image Denoising Algorithms, with a New One
Example-based super-resolution
Fast and robust multiframe super resolution
Improving resolution by image registration
Learning Low-Level Vision
Related Papers (5)
Frequently Asked Questions (12)
Q2. Why do small patches recur in multiple scales of the same image?
due to the perspective projection of cameras, images tend to contain scene-specific information in diminishing sizes (diminishing toward the horizon), thus recurring in multiple scales of the same image.
Q3. What is the problem of recovering H?
In principle, when there is only a single low-resolution image L = ( H ∗ B ) ↓s, the problem of recovering H becomes under-determined, as the number of constraints induced by L is smaller than the number of unknowns in H .
Q4. How many times do patches recur in the same image?
Most of them recur several times within and across scales of the same image (more than 80% of the patches recur 9 or more times in the original image scale; more than 70% recur 9 or more times at 0.41 of the input scale, and 60% of them recur 9 or more times in 0.26 of the input scale.)
Q5. What is the reason for the recurrence of patch repetitions in images?
when much smaller imagepatches are used, e.g., 5 × 5, such patch repetitions occur abundantly within and across image scales, even when the authors do not visually perceive any obvious repetitive structure in the image.
Q6. How many pixel neighbors are there in the same image?
For each pixel in L find its k nearest patch neighbors in the same image L (e.g., using an Approximate Nearest Neighbor algorithm [1]; the authors typically use k=9) and compute their subpixel alignment (at 1s pixel shifts, where s is the scale factor.)
Q7. What are the constraints used to solve the equations for image Il?
When solving the equations for image Il+1, the authors employed not only the low-res/high-res patch correspondences found in the input image L, but also all newly learned patch correspondences from the newly recovered high-res images so far: I0, ..., Il.
Q8. What is the main improvement in resolution from the Example-Based SR component?
the Classical-SR component (apart from providing small resolution increase - see Fig. 5c), plays a central role in preventing the Example-Based SR component from hallucinating erroneous high-res details (a problem alluded to by [11]).
Q9. What is the definition of a patch recurrence?
An input patch “recurs” in another scale if it appears ‘as is’ (without blurring, subsampling, or scaling down) in a scaled-down version of the image.
Q10. What is the corresponding patch in the low-res image?
Then its higher-res ‘parent’ patch, Q0(sl · p̃), can be extracted from the input image I0 = L (or from any intermediate resolution level between I−l and L, if desired).
Q11. What scales give rise to ExampleBased SR?
Recurrence of patches across different scales gives rise to ExampleBased SR from a single image, with no prior examples (Sec. 3.2).
Q12. What is the recurrence of patches in the same image scale?
Recurrence of patches within the same image scale forms the basis for applying the Classical SR constraints to information from a single image (Sec. 3.1).