SLIC Superpixels Compared to State-of-the-Art Superpixel Methods
read more
Citations
CGIntrinsics: Better Intrinsic Image Decomposition through Physically-Based Rendering
Point Cloud Oversegmentation With Graph-Structured Deep Metric Learning
Deep convolutional networks for pancreas segmentation in CT imaging
Supervised Evaluation of Image Segmentation and Object Proposal Techniques
Correspondence Driven Saliency Transfer
References
The Pascal Visual Object Classes (VOC) Challenge
Normalized cuts and image segmentation
Least squares quantization in PCM
Normalized cuts and image segmentation
Mean shift: a robust approach toward feature space analysis
Related Papers (5)
Frequently Asked Questions (14)
Q2. What is the efficient method for storing distances between pixel centers?
SLIC is the most memory efficient method, requiring only N floats to store the distance from each pixel to its nearest cluster center.
Q3. What is the L2 norm used to compute?
The L2 norm is used to compute a residual error E between the new cluster center locations and previous cluster center locations.
Q4. What is the way to search for similar pixels?
Since the expected spatial extent of a superpixel is a region of approximate size S × S, the search for similar pixels is done in a region 2S × 2S around the superpixel center.
Q5. What is the common method for generating superpixels?
Starting from a rough initial clustering of pixels, gradient ascent methods iteratively refine the clusters until some convergence criterion is met to form superpixels.
Q6. What is the way to generate superpixels?
It performs an agglomerative clustering of pixels as nodes on a graph, such that each superpixel is the minimum spanning tree of the constituent pixels.
Q7. What is the main reason for the compactness of superpixels?
regular superpixels are often desirable because their bounded size and few neighbors form a more interpretable graph and can extract more locally relevant features.
Q8. What is the key to speeding up the clustering algorithm?
2. This is the key to speeding up their algorithm because limiting the size of the search region significantly reduces the number of distance calculations, and results in a significant speed advantage over conventional kmeans clustering where each pixel must be compared with all cluster centers.
Q9. How many iterations can be used to find the closest cluster center?
The assignment and update steps can be repeated iteratively until the error converges, but the authors have found that 10 iterations suffices for most images, and report all results in this paper using this criteria.
Q10. What is the important property of a superpixel method?
Examples of superpixel segmentations produced by each method appear in Fig. 7.Arguably, the most important property of a superpixel method is its ability to adhere to image boundaries.
Q11. What is the method for generating superpixels?
The authors propose a new method for generating superpixels which is faster than existing methods, more memory efficient,exhibits state-of-the-art boundary adherence, and improves the performance of segmentation algorithms.
Q12. What is the cost of reducing the resolution of the image?
For some applications, such as mitochondria segmentation from electron micrographs (EM), the images are large but reducing the resolution is not an option.
Q13. What is the efficient method for storing edge weights and thresholds?
Other methods have comparatively high memory requirements: GS04 and GC10 require 5N floats to store edge weights and thresholds for 4-connectivity (or 9N for 8-connectivity).
Q14. How can the authors avoid the problem of defining distances?
This problem can be avoided by fixing Nc to a constant m so that Eq. 1 becomesD′ = √( dc m )2 + ( ds S )2 , (2)which simplifies to the distance measure the authors use in practiceD = √ dc 2 + ( ds S )2 m2. (3)By defining D in this manner, m also allows us to weigh the relative importance between color similarity and spatial proximity.