Project topics¶
Each group (with two students) needs to choose one project from the following topics. Please read the project guideline, which gives details regarding how to implement, write and submit your project-documents.
Topic 1: Deconvolution using Richardson-Lucy algorithm¶
This project is about blind deconvolution based on the Richardson–Lucy deconvolution algorithm. The paper presents a method for accurate deconvolutions of a quality near that of a deconvolution with full knowledge of the PSF. Read project description here.
Topic 2: Non-Local means for image denoising¶
Denoising is one of the oldest topics in image restoration. NL-means is a denoisng approach based on a nonlocal averaging of all pixels in the image. Read project description here.
Topic 3: Image inpainting¶
For different reasons, you might need to remove large part of a digital image. The challenge is to fill in properly the missing information or to “inpaint” it. This project aims to do such a task. Read project description here.
Topic 4: Edge-aware image editing¶
Edges are important features in digital images. They define the boundaries of objects and divide the image into separate semantic sub-regions. However, many image processing algorithms fail to preserve this important feature. Based on a modified Laplacian pyramid, this paper presents a new approach for tone mapping and image abstraction. Read project description here.
Topic 5: Measuring image blur/sharpness¶
Quantifying the image blur/sharpness can be useful in certain application such subject detection, no-reference image quality assessment and image autoenhancement. This paper presents an algorithm for measuring the local perceived sharpness in an image. Read project description here.
Topic 6: Image fusion using wavelet transform¶
Image fusion is a ubiquitous technique in many different applications such as remote sensing, medical imaging and military surveillance applications. The aim of image fusion is to combine image information that are acquired using different modalities into a single reconstruction. This paper is one of the earliest attempt on this topic which makes use of wavelet transform for this purpose. Read project description here.
Topic 7: Digital extended depth of focus¶
Focus fusion is an attempt to fuse the stack of images of the same scene with different focus/defocus properties. This is very similar to image fusion, however, here the images are typically from the same modalities. In this project, you need to implement two methods for focus fusion. Read project description here.
Topic 8: Optic flow¶
Optic flow seeks to find the motion field between two frames (consecutive images) of an image series like video. The method of Lucas and Kanade is one of the very first methods addressing this problem by minimizing an L_2 norm measure. Read project description here.
Topic 9: Particle averaging¶
Single molecule localization is a relatively new light microscopy technique which allows imaging below the diffraction limit. However, due to practical difficulties, the achievable resolution is still limited to about 20 nm. When multiple copies of the same structure (“particle”) are available, proper alignment of these particles will improve resolution. Following your template matching assignment (in labwork 4), you are going to align and average multiple instances of NPCs to achieve higher image resolution. Read project description here.
Topic 10: Video magnification¶
Many seemingly static scenes contain subtle changes that are invisible to the naked human eye. It is possible to pull out these small changes from videos through the use of video magnification algorithms. Those algorithms give a way to visualize these small changes by amplifying them. Examples include: blood ow in a face, the human pulse and the motion of hot air. In this project you will implement such methods. We ask you to make a report and a presentation that demonstrates understanding, insightful and relevant comparisons. You will run existing Matlab code, record some videos, and implement a video magnification method yourself. Read project description here.
Topic 11: Poisson Image Editing¶
For various reasons, you might need to edit an image desirably without noticeable effect. Poisson image editing is a framework with which different image editing tasks such as seamless image cloning, texture flattening and local illumination enhancement can be implemented efficiently. The most important feature of this approach is that it realizes such techniques on image gradient rather than pixel intensity domain. Read project description here.
Topic 12: Color transfer¶
Color transfer refers to an image manipulation technique in which the palette of an example target picture is being mapped to another picture. In this project, you are going to implement a color transfer technique which tries to find a differentiable mapping that transforms the original color probability density function (PDF) to a new target color PDF. Read project description here.
Topic 13: Level sets¶
At the basis of many image processing algorithms is a segmentation of the object of interest. This allows for analysis of shape and volume features. In liver surgery, for instance, it is important to know the size of the liver. In this project we want you to segment the liver through a very sophisticated technique: level sets. For more information, please read the project description here.
Topic 14: Compressed sensing MRI¶
Compressed sensing (CS) aims to reconstruct signals and images from undersampled measurements. Until recently, this was considered not possible. Magnetic resonance imaging (MRI) is an medical imaging technique with a slow data acquisition process. Compressed sensing MRI potentially reduces scan time which is beneficial for both patients and health care economics. You can read the description of this project here
Topic 15: Linear discriminant analysis for recognising differences¶
One might expect that differences occur between images from different groups of subjects, e.g. men versus women, old versus young, normals versus diseased. However, it is often not so trivial to pinpoint exactly where these differences are. Linear discriminant analysis is a method to identify in which way the data from one group changes in the data from another group. In this project we want you to implement this technique to study differences between two groups of subjects. You can read the description of this project here.