The significance of image compression in plant phenotyping applications

Abstract

We currently witness an increasingly higher throughput in image-based plant phenotyping experiments. The majority of imaging data are collected based on complex automated procedures, and are then post-processed to extract phenotyping related information. In this article we show that image compression used in such procedures may compromise phenotyping results and needs to be taken into account. We motivate the paper with three illuminating proof of concept experiments which demonstrate that compression (especially in its most common lossy form of JPEG) does affect measurements of plant traits and errors introduced can be high. We further systematically explore how compression affects measurement fidelity, quantified as effects on image quality as well as errors in extracted plant visual traits. To do so we evaluate a variety of image-based phenotyping scenarios, including size and color of shoots, leaf and root growth. To show that even visual impression can be used to assess compression effects we use root system images as examples. Overall, we find that compression has a considerable effect on several types of analyses (albeit visual or quantitative) and that proper care is necessary to ensure that such choice does not affect biological findings. In order to avoid or at least minimize introduced measurement errors, for each scenario we derive recommendations and provide guidelines on how to identify suitable compression options in practice. We also find that certain compression choices can offer beneficial returns, in terms of reducing the amount of data storage without compromising phenotyping results. This may enable even higher throughput experiments in the future.

Phenotiki: an open software and hardware platform for affordable and easy image‐based phenotyping of rosette‐shaped plants

“…Furthermore, by relying on the Cloud, the additional computational needs to analyze higher throughput data can be readily met because of its immediate resource scalability, and the implementation of asynchronous upload mechanisms that are used by our device to send data to the Cloud. When the available network bandwidth or storage capacity is limited (which could occur in laboratories in countries with poorer Internet infrastructure), we can potentially integrate image compression algorithms within the Phenotiki device (Minervini and Tsaftaris, ; Minervini et al ., ).…”

“…To facilitate the configuration and monitoring of the device, we deployed a web‐based graphical user interface to operate it remotely from a laptop or a smartphone (Figures b and S2; Video clip S1). To reduce storage requirements without affecting phenotyping accuracy (Minervini et al ., ), images were encoded at the device using the lossless compression standard available in the PNG file format (although Phenotiki supports a variety of lossless and lossy image formats). At the end of the experiment, a ZIP archive containing all of the images acquired was automatically created on the Phenotiki device, and via the web‐based interface of the device we downloaded it to a local workstation for storage and processing (Figure S2).…”

Accuracy of image analysis tools for functional root traits: A comment on Delory et al. (2017)

“…WinRhizo ™ , appears to choose the threshold correctly, at least if images have a clear contrast, whereas IJ_Rhizo introduced large errors by misclassification. Our images contained minor variability in colour intensity (Supplement 3) as a result of image conversion (Minervini, Scharr, & Tsaftaris, ), which, nevertheless, introduced directional pixel misclassification towards less pixels being classified as root by IJ_Rhizo at 1,200 dpi but not 800 dpi (Supplement 3, zoom to see the effect). This explains the great overestimation of root length and underestimation of diameters we observed with automatic threshold at 1,200 dpi, since single root segments are partly interpreted as multiple very thin segments by IJ_Rhizo.…”

Discrete cosine transform based processing framework for indexing, decomposition and compression of biospeckle data

“…However, in practice, such data storage typically requires a lot of virtual memory resources. Importance of compression has also increased significantly due to online storage, sharing and transmission of images in different research areas (medical, agriculture, biomedical, engineering, biology and many more) [22][23][24]. In the field of medical imaging large number of images are produced for research, surgical applications and disease diagnostics.…”