Optimizing and Learning for Super-resolution Lyndsey C. Pickup, Stephen J. Roberts & Andrew...

40
Optimizing and Learning for Super-resolution Lyndsey C. Pickup, Stephen J. Roberts & Andrew Zisserman Robotics Research Group, University of Oxford

Transcript of Optimizing and Learning for Super-resolution Lyndsey C. Pickup, Stephen J. Roberts & Andrew...

Page 1: Optimizing and Learning for Super-resolution Lyndsey C. Pickup, Stephen J. Roberts & Andrew Zisserman Robotics Research Group, University of Oxford.

Optimizing and Learning forSuper-resolution

Lyndsey C. Pickup, Stephen J. Roberts

& Andrew Zisserman

Robotics Research Group, University of Oxford

Page 2: Optimizing and Learning for Super-resolution Lyndsey C. Pickup, Stephen J. Roberts & Andrew Zisserman Robotics Research Group, University of Oxford.

The Super-resolution Problem

Given a number of low-resolution imagesdiffering in: geometric transformations lighting (photometric) transformations camera blur (point-spread function) image quantization and noise.

Estimate a high-resolution image:

Page 3: Optimizing and Learning for Super-resolution Lyndsey C. Pickup, Stephen J. Roberts & Andrew Zisserman Robotics Research Group, University of Oxford.

Low-resolution image 1

Page 4: Optimizing and Learning for Super-resolution Lyndsey C. Pickup, Stephen J. Roberts & Andrew Zisserman Robotics Research Group, University of Oxford.

Low-resolution image 2

Page 5: Optimizing and Learning for Super-resolution Lyndsey C. Pickup, Stephen J. Roberts & Andrew Zisserman Robotics Research Group, University of Oxford.

Low-resolution image 3

Page 6: Optimizing and Learning for Super-resolution Lyndsey C. Pickup, Stephen J. Roberts & Andrew Zisserman Robotics Research Group, University of Oxford.

Low-resolution image 4

Page 7: Optimizing and Learning for Super-resolution Lyndsey C. Pickup, Stephen J. Roberts & Andrew Zisserman Robotics Research Group, University of Oxford.

Low-resolution image 5

Page 8: Optimizing and Learning for Super-resolution Lyndsey C. Pickup, Stephen J. Roberts & Andrew Zisserman Robotics Research Group, University of Oxford.

Low-resolution image 6

Page 9: Optimizing and Learning for Super-resolution Lyndsey C. Pickup, Stephen J. Roberts & Andrew Zisserman Robotics Research Group, University of Oxford.

Low-resolution image 7

Page 10: Optimizing and Learning for Super-resolution Lyndsey C. Pickup, Stephen J. Roberts & Andrew Zisserman Robotics Research Group, University of Oxford.

Low-resolution image 8

Page 11: Optimizing and Learning for Super-resolution Lyndsey C. Pickup, Stephen J. Roberts & Andrew Zisserman Robotics Research Group, University of Oxford.

Low-resolution image 9

Page 12: Optimizing and Learning for Super-resolution Lyndsey C. Pickup, Stephen J. Roberts & Andrew Zisserman Robotics Research Group, University of Oxford.

Low-resolution image 10

Page 13: Optimizing and Learning for Super-resolution Lyndsey C. Pickup, Stephen J. Roberts & Andrew Zisserman Robotics Research Group, University of Oxford.

Super-Resolution Image

Page 14: Optimizing and Learning for Super-resolution Lyndsey C. Pickup, Stephen J. Roberts & Andrew Zisserman Robotics Research Group, University of Oxford.

Generative Model

Registrations, lighting and

blur.

High-resolution image, x.

y1 y2 y3 y4

Low-resolution images

W4W3W2W1

Page 15: Optimizing and Learning for Super-resolution Lyndsey C. Pickup, Stephen J. Roberts & Andrew Zisserman Robotics Research Group, University of Oxford.

Generative Model

• Geometric registrations• Point-spread function• Photometric registrations

We don’t have:We have:• Set of low-resolution input images, y.

Page 16: Optimizing and Learning for Super-resolution Lyndsey C. Pickup, Stephen J. Roberts & Andrew Zisserman Robotics Research Group, University of Oxford.

Maximum a Posteriori (MAP) Solution

Standard method:

1. Compute registrations from low-res images. 2. Solve for SR image, x, using gradient descent.

y1 y2 y3 y4

W4W3W2

W1

x

[Irani & Peleg ‘90, Capel ’01, Baker & Kanade ’02, Borman ‘04]

Page 17: Optimizing and Learning for Super-resolution Lyndsey C. Pickup, Stephen J. Roberts & Andrew Zisserman Robotics Research Group, University of Oxford.

What’s new

1. We solve for registrations and SR image jointly.

2. We also find appropriate values for parameters in the prior term at the same time.

Hardie et al. ’97: adjust registration while optimizing super-resolution estimate.

• Exhaustive search limits them to translation only. • Simple smoothness prior softens image edges.

i.e. given the low-res images, y, we solve for the SR image x and the mappings, W simultaneously.

y1 y2 y3 y4

W4W3W2

W1

x

Page 18: Optimizing and Learning for Super-resolution Lyndsey C. Pickup, Stephen J. Roberts & Andrew Zisserman Robotics Research Group, University of Oxford.

Overview of rest of talk

• Simultaneous Approach– Model details

– Initialisation technique

– Optimization loop

• Learning values for the prior’s parameters

• Results on real images

Page 19: Optimizing and Learning for Super-resolution Lyndsey C. Pickup, Stephen J. Roberts & Andrew Zisserman Robotics Research Group, University of Oxford.

Maximum a Posteriori (MAP) Solution

Image x. Corrupt with additive Gaussian noise.

Warp, with parameters Φ.

Blur by point-spread function.

Decimate by zoom factor.

y1 y2 y3 y4

W4W3W2

W1

x

y

Page 20: Optimizing and Learning for Super-resolution Lyndsey C. Pickup, Stephen J. Roberts & Andrew Zisserman Robotics Research Group, University of Oxford.

Details of Huber Prior

Huber function is quadratic in the middle, and linear in the tails.

Probability distribution is like a heavy-tailed Gaussian.

ρ (z,α) p (z|α,v)

Red: large αBlue: small α

This is applied to image gradients in the SR image estimate.

Page 21: Optimizing and Learning for Super-resolution Lyndsey C. Pickup, Stephen J. Roberts & Andrew Zisserman Robotics Research Group, University of Oxford.

Details of Huber Prior

Ground Truth

α=0.1 v=0.4

Too little smoothingToo much smoothing

α=0.05 v=0.05 α=0.01 v=0.01 α=0.01 v=0.005

Edges are sharper

Advantages: simple, edge-preserving, leads to convex form for MAP equations.

Solutions as α and v vary:

Page 22: Optimizing and Learning for Super-resolution Lyndsey C. Pickup, Stephen J. Roberts & Andrew Zisserman Robotics Research Group, University of Oxford.

Advantages of Simultaneous Approach

Learn from lessons of Bundle Adjustment: improve results by optimizing the scene estimate and the registration together.

Registration can be guided by the super-resolution model, not by errors measured between warped, noisy low-resolution images.

Use a non-Gaussian prior which helps to preserve edges in the super-resolution image.

Page 23: Optimizing and Learning for Super-resolution Lyndsey C. Pickup, Stephen J. Roberts & Andrew Zisserman Robotics Research Group, University of Oxford.

Overview of Simultaneous Approach

1. Start from a feature-based RANSAC-like registration between low-res frames.

2. Select blur kernel, then use average image method to initialise registrations and SR image.

3. Iterative loop: Update Prior Values Update SR estimate Update registration estimate

Page 24: Optimizing and Learning for Super-resolution Lyndsey C. Pickup, Stephen J. Roberts & Andrew Zisserman Robotics Research Group, University of Oxford.

Use average image as an estimate of the super-resolution image (see paper).

Minimize the error between the average image and the low-resolution image set.

Use an early-stopped iterative ML estimate of the SR image to sharpen up this initial estimate.

Initialisation

Average image

ML-sharpened estimate

Page 25: Optimizing and Learning for Super-resolution Lyndsey C. Pickup, Stephen J. Roberts & Andrew Zisserman Robotics Research Group, University of Oxford.

1.Update prior’s parameter values (next section)

2.Update estimate of SR image

3.Update estimate of registration and lighting values, which parameterize W

Repeat till converged.

Optimization Loop

Page 26: Optimizing and Learning for Super-resolution Lyndsey C. Pickup, Stephen J. Roberts & Andrew Zisserman Robotics Research Group, University of Oxford.

Joint MAP ResultsD

ecre

asin

g p

rio

r st

ren

gth

Registration Fixed Joint MAP

Page 27: Optimizing and Learning for Super-resolution Lyndsey C. Pickup, Stephen J. Roberts & Andrew Zisserman Robotics Research Group, University of Oxford.

Learning Prior Parameters α, ν

Split the low-res images into two sets:

Use first set to obtain an SR image.

Find error on validation set.

Page 28: Optimizing and Learning for Super-resolution Lyndsey C. Pickup, Stephen J. Roberts & Andrew Zisserman Robotics Research Group, University of Oxford.

Learning Prior Parameters α, ν

Split the low-res images into two sets:

Use first set to obtain an SR image.

Find error on validation set.

But what if one of the validation images is mis-registered?

Page 29: Optimizing and Learning for Super-resolution Lyndsey C. Pickup, Stephen J. Roberts & Andrew Zisserman Robotics Research Group, University of Oxford.

Learning Prior Parameters α, ν

Instead, we select pixels from across all images, choosing differently at each iteration.

We evaluate an SR estimate using the unmarked pixels, then use the forward model to compare the estimate to the red pixels.

Page 30: Optimizing and Learning for Super-resolution Lyndsey C. Pickup, Stephen J. Roberts & Andrew Zisserman Robotics Research Group, University of Oxford.

Learning Prior Parameters α, ν

Instead, we select pixels from across all images, choosing differently at each iteration.

We evaluate an SR estimate using the unmarked pixels, then use the forward model to compare the estimate to the red pixels.

Page 31: Optimizing and Learning for Super-resolution Lyndsey C. Pickup, Stephen J. Roberts & Andrew Zisserman Robotics Research Group, University of Oxford.

Learning Prior Parameters α, ν

To update the prior parameters:

1. Re-select a cross-validation pixel set.

2. Run the super-resolution image MAP solver for a small number of iterations, starting from the current SR estimate.

3. Predict the low-resolution pixels of the validation set, and measure error.

4. Use gradient descent to minimise the error with respect to the prior parameters.

Page 32: Optimizing and Learning for Super-resolution Lyndsey C. Pickup, Stephen J. Roberts & Andrew Zisserman Robotics Research Group, University of Oxford.

Results: Eye Chart

MAP version: fixing registrations then

super-resolving

Joint MAP version with adaptation of prior’s

parameter values

Page 33: Optimizing and Learning for Super-resolution Lyndsey C. Pickup, Stephen J. Roberts & Andrew Zisserman Robotics Research Group, University of Oxford.

Results: Groundhog Day

Page 34: Optimizing and Learning for Super-resolution Lyndsey C. Pickup, Stephen J. Roberts & Andrew Zisserman Robotics Research Group, University of Oxford.

The blur estimate can still be altered to change the SR result. More ringing and artefacts appear in the regular MAP version.

Results: Groundhog Day

Blur radius = 1 Blur radius = 1.4 Blur radius = 1.8

Regular MAP

Simultaneous

Page 35: Optimizing and Learning for Super-resolution Lyndsey C. Pickup, Stephen J. Roberts & Andrew Zisserman Robotics Research Group, University of Oxford.

Lola Rennt

Page 36: Optimizing and Learning for Super-resolution Lyndsey C. Pickup, Stephen J. Roberts & Andrew Zisserman Robotics Research Group, University of Oxford.

Real Data: Lola Rentt

Page 37: Optimizing and Learning for Super-resolution Lyndsey C. Pickup, Stephen J. Roberts & Andrew Zisserman Robotics Research Group, University of Oxford.

Real Data: Lola Rentt

Page 38: Optimizing and Learning for Super-resolution Lyndsey C. Pickup, Stephen J. Roberts & Andrew Zisserman Robotics Research Group, University of Oxford.

Real Data: Lola Rentt

Page 39: Optimizing and Learning for Super-resolution Lyndsey C. Pickup, Stephen J. Roberts & Andrew Zisserman Robotics Research Group, University of Oxford.

Real Data: Lola Rentt

Page 40: Optimizing and Learning for Super-resolution Lyndsey C. Pickup, Stephen J. Roberts & Andrew Zisserman Robotics Research Group, University of Oxford.

Conclusions

• Joint MAP solution: better results by optimizing SR image and registration parameters simultaneously.

• Learning prior values: preserve image edges without having to estimate image statistics in advance.

• DVDs: Automatically zoom in on regions with a registrations up to a projective transform and with an affine lighting model.

• Further work: marginalize over the registration – see NIPS 2006.