Lossy refers to a compression scheme wherein the image is close to the original image after its is decompressed, but not identical. In other words, some information is lost during the compression procedure.
An example is the JPEG (Joint Photographic Experts Group) format.
An example is the JPEG (Joint Photographic Experts Group) format.
The advantage of this type of compression is the ability to compress the original image to a much smaller size for storage (i.e. much larger compression ratio) compared to a lossless format, but usually at the cost of quality or lost information. Thus, compression ratio and image quality have an inverse relationship.
It manages larger compression ratios than lossless compression by taking advantage of redundancies in the data encoding for a particular image, which can generally be categorized as: psychometric, encoding, or inter-pixel redundancies. For example, the inability for humans to discern between two very similar levels of grey would allow the image to save space by using one level for both.
As a result of psychometric manipulations, image quality post-compression is difficult to measure quantitatively. There exist various methods for measuring image quality, but the Q index developed by Wang et al has shown good correlation with human visual assessments of quality (Shiao, 2007).
It manages larger compression ratios than lossless compression by taking advantage of redundancies in the data encoding for a particular image, which can generally be categorized as: psychometric, encoding, or inter-pixel redundancies. For example, the inability for humans to discern between two very similar levels of grey would allow the image to save space by using one level for both.
As a result of psychometric manipulations, image quality post-compression is difficult to measure quantitatively. There exist various methods for measuring image quality, but the Q index developed by Wang et al has shown good correlation with human visual assessments of quality (Shiao, 2007).
The typical compression ratio for a JPEG image is between 10 and 100 (Williams) with . Such a wide range is simply due to a unique aspect of the JPEG compression scheme, which gives the ability to vary the level of compression at the cost of image quality. Therefore, depending on your needs, you can opt for a higher compression ratio but lower quality, or a lower compression ratio but higher quality. However, it is important to keep in mind that compression is irreversible, meaning once information is lost from an image it cannot be recovered.
Based on an analysis of papers from 1997-2004, researchers found most studies suggested a compression ratio of 15:1 was the ideal for lossy compression schemes in a medical/radiological setting (Shiao et al., 2007). Although this depends on the imaging modality (e.g. CT vs. MRI), anatomical feature (e.g. mammogram vs. brain), and particular imaging scheme (e.g. JPEG vs. JPEG-2000). For instance, a recent study found compression ratios of up to 60:1 may be acceptable for use in medical diagnostics of JPEG-2000 encoded mammograms (Kang et al., 2011).
Similarly, more formats that would allow for higher compression ratios are highly sought after, as increasing digital information is making storage and bandwidth harder to come by. However, it cannot be at the expense of image accuracy, as higher compression ratios often lead to distortions in the image referred to as compression artefacts. These distortions can be especially disastrous in medical images, where it is often little differences/distortions that are the difference between normal and pathological states, thereby leading to an increase in false positives and negatives.
Based on an analysis of papers from 1997-2004, researchers found most studies suggested a compression ratio of 15:1 was the ideal for lossy compression schemes in a medical/radiological setting (Shiao et al., 2007). Although this depends on the imaging modality (e.g. CT vs. MRI), anatomical feature (e.g. mammogram vs. brain), and particular imaging scheme (e.g. JPEG vs. JPEG-2000). For instance, a recent study found compression ratios of up to 60:1 may be acceptable for use in medical diagnostics of JPEG-2000 encoded mammograms (Kang et al., 2011).
Similarly, more formats that would allow for higher compression ratios are highly sought after, as increasing digital information is making storage and bandwidth harder to come by. However, it cannot be at the expense of image accuracy, as higher compression ratios often lead to distortions in the image referred to as compression artefacts. These distortions can be especially disastrous in medical images, where it is often little differences/distortions that are the difference between normal and pathological states, thereby leading to an increase in false positives and negatives.
As an example, the algorithm for JPEG compression makes it particularly prone to blocking artefacts at higher compression ratios. Basically, each 8 x 8 set of pixels is analysed as a separate block and quantified independently, which leads to boundaries between the blocks that become more apparent at higher compression ratios.
|