Oct 12, 2018
Lopata Hall, Room 101
"Identifying Recurring Patterns with DNN for Natural Image Denoising"
Adviser: Ayan Chakrabarti
While there is a vast diversity in the patterns and textures that occur across different varieties of natural images, the variance of such patterns within a single image is far more limited. A variety of traditional methods have exploited this self-similarity or recurrence with considerable success for image modeling, estimation, and restoration. A key challenge, however, is in accurately identifying recurring patterns within degraded image observations. In this talk, I will present a new method for natural image denoising, that trains a deep neural network to determine whether noisy patches share common underlying patterns. Specifically, given a pair of noisy patches, the network predicts whether different transform sub-band coefficients of the original noise-free patches are the same. The denoising algorithm averages these matched coefficients to obtain an initial estimate of the clean image, with much higher quality than traditional approaches. This estimate is then refined with a second post-processing network, yielding state-of-the-art denoising performance.
"Plants have Skeletons too: Using Medial Axes for Plant Shape Analysis"
Adviser: Tao Ju
The medial axes skeleton is defined as the set of all points of a shape which have more than one closest point on the object’s boundary. By being a lower-dimensional, thinned representation of the original shape, the medial axes concisely describes a shape’s geometry and topology. This gives rise to applications of the medial axes in biomedicine, plant biology, animation, and many other areas. We present novel algorithms and applications of the medial axes in analyzing plant architecture, an area of plant biology which has a huge impact on crop production and survivability. My focus will be on medial axes-based branch and bract segmentation algorithms for panicles of sorghum, the fifth most widely produced crop in the world. By combining properties of the medial axes with domain-specific knowledge of plant architecture, our approach offers significant geometric, computational, and topological advantages over existing ones both in theory and practice. This plants the seed for other applications of the medial axes in plant biology, such as in analyzing root system architecture.