What Full-Reference Quality Metrics Reveal About Verifying a Claim of Preserved Detail
Every standard method the video and audio industries actually use to verify that a compressed file preserved quality, VMAF, SSIM, PSNR for video, PEAQ, POLQA, PESQ for audio, requires the original to compare against. This series has used a claim of preserved quality at reduced size four separate times, for video, for audio, and now for general web content, without ever independently verifying it. This paper examines the claim directly, on its own, rather than as a supporting detail for something else.
Three earlier papers in this series, on cinema, on live sports and entertainment delivery, and on spatial audio rendering, each described a resolution-independent or bandwidth-independent quality claim in supporting detail, then flagged the same limitation each time: the claim is drawn from MindAptiv's own public materials, not independently tested by this series. This paper takes up that claim directly, as its own subject, and adds a fourth instance found since: Jewel®, MindAptiv's browser layer, applying the identical claim to general web payload rather than video or audio specifically.
This paper applies Signal Paper I's doctrine, Captured ≠ Governed, to the claim itself: a smaller file that looks the same is not a file whose fidelity has been proven. It grounds that argument in the real, standardized, full-reference quality metrics the video and audio industries actually use, VMAF, SSIM, and PSNR for video, PEAQ, POLQA, and PESQ for audio, all of which require the original to verify against, and states precisely what MindAptiv's own patents do and do not establish about the underlying mechanism.
A compressed video, audio file, or web page that looks and sounds correct to a casual viewer or listener has cleared the lowest bar a quality claim can clear: nothing looks obviously broken. Every real, standardized method the video and audio industries use to make a stronger claim than that, that a specific compression method preserved a measurable, comparable degree of fidelity, requires something a casual viewing does not: the original, uncompressed source, held alongside the compressed output, so the two can be measured against each other directly.
This series has invoked a version of this claim four times: in Signal Paper X, describing a resolution-independent rendering technique for cinema placements; in Signal Paper XI, describing WarpSpeed™'s bandwidth-independent video fidelity at 28 kbps; in Signal Paper XII, extending the same mechanism to spatial audio by analogy; and now, found since, in Jewel®'s own materials, describing the identical 28 kbps demonstration applied to general web payload. Each of the first three instances carried the same caveat: drawn from MindAptiv's own characterization, not independently tested. This paper asks what independent testing would actually require, and states plainly what does and does not currently exist to satisfy it.
The video industry's standard tools for measuring whether a compressed file preserved quality are Peak Signal-to-Noise Ratio (PSNR), the Structural Similarity Index (SSIM), and Video Multi-Method Assessment Fusion (VMAF), developed by Netflix and introduced in 2016. PSNR is a direct, pixel-level numerical comparison; SSIM measures structural similarity in luminance, contrast, and structure; VMAF is a machine-learning model trained on human-rated video clips, fusing several lower-level metrics into a single score designed to correlate with actual human perception. All three, without exception, are full-reference: each one works by comparing the compressed output directly against the original, uncompressed source. None of them can be run on the compressed file alone.
The audio industry's equivalent standards follow the identical pattern. The Perceptual Evaluation of Audio Quality (PEAQ), standardized as ITU-R Recommendation BS.1387 in 1998 and last revised in 2023, was built specifically to grade perceptual audio codecs like MP3 and AAC, and outputs an Objective Difference Grade from 0, imperceptible from the original, down to -4, very annoying. The Perceptual Objective Listening Quality Analysis (POLQA, ITU-T P.863) and its predecessor PESQ (ITU-T P.862) serve the same function for speech and voice codecs. Every one of these is also full-reference. Reference-free ("no-reference") quality assessment exists as an active area of research, but published comparisons of the two approaches consistently find that full-reference methods correlate better with actual human judgment, which is precisely why full-reference remains the industry's working standard rather than a stopgap awaiting replacement.
The underlying mechanism behind MindAptiv's resolution-independent rendering claim is more specifically documented than this series has previously stated. U.S. Patent 10,037,592 (2018) describes computing first- and second-order gradients of a two-dimensional image signal, representing them as quaternions, and using the logarithm of those quaternions to determine gradient magnitude and orientation, information used to construct an output signal with greater detail than the input. U.S. Patent 10,846,821 (2020) applies the identical gradient technique explicitly to video signals rather than static images. U.S. Patent 11,373,272 (2022) extends the same family to static images, moving images, three-dimensional images, and other data types, and separately names, among its stated field of invention, natural language interfaces and improved and customizable web browsers. MindAptiv's own materials describe this gradient-and-logarithm technique as the mathematical basis of Nebulo®'s level-of-detail traversal, the same mechanism this series has referenced, without this specific grounding, in Signal Papers X, XI, and XII.
This is a real, verifiable, and more specific technical foundation than "MindAptiv's own characterization" implies on its own. It is not, however, independent verification that the technique achieves the quality-preservation results claimed for any specific deployment. A patent describes what a method computes and claims as its intended effect; it does not, and is not designed to, independently confirm the method's real-world output against a specific full-reference metric on specific content. That confirmation is a separate, empirical question, and it is the one this paper is actually about. There is also a basic information-theoretic reason this question can never be waved away entirely: rate-distortion theory, formalized by Claude Shannon in 1959, establishes that below a certain data rate, some loss of information relative to the original is mathematically unavoidable for any compression method, not a limitation specific to any one company's implementation. A "no loss of detail" claim, taken completely literally, is a claim about which specific losses are imperceptible or immaterial, not a claim that no information was discarded at all.
The architectural basis for extending this claim to video, audio, and general web payload follows the same patent scope discussed here and established more generally in Signal Paper I. This paper does not re-derive that claim or its stated limits; see Signal Paper I, Section 05, for what has and has not been independently reviewed in the patents' full claim language, a limit that applies here as well: this paper reviewed the patents' titles, abstracts, and specification text as quoted in MindAptiv's own patent-assessment materials, not the complete prosecution history or full claim set of any of the three patents.
The same underlying claim, that a MindAptiv mechanism preserves perceptual quality while substantially reducing size or bandwidth, has now appeared four times across this series, in four different signal types, each time without an accompanying full-reference metric score, an independent test, or a governed record of either.
Each instance is architecturally consistent with the same underlying mechanism, and each rests on the same evidentiary footing: a real, patent-documented technique, described by MindAptiv in its own materials, without an independently computed VMAF, SSIM, PEAQ, or comparable score attached to any specific deployment cited in this series.
This paper does not claim that MindAptiv's resolution- or bandwidth-independent rendering technique fails to preserve quality; it claims that no independent, full-reference quality score has been cited anywhere in this series to confirm that it does, for any of the four instances described in Section 04. It does not claim that the underlying patents are invalid, narrow, or unusual; their field-of-invention language, as quoted in MindAptiv's own materials, is specific and appears to cover image, video, three-dimensional, and browser-related signal processing directly. It does not claim to have reviewed the patents' full prosecution history or complete claim set, only their titles, abstracts, and specification text as quoted in a MindAptiv-published assessment page, a limit stated in Section 03 as well.
This vertical inherits a pattern this series has used before, taking a claim that appeared as supporting detail elsewhere and examining it as a subject in its own right, the way Signal Paper III took a market-sizing figure seriously enough to trace its actual sourcing. What distinguishes this paper is that the claim under examination is MindAptiv's own, repeated across four instances without ever being independently tested, rather than an external claim this series checked against outside sources.
It follows Emergency Medical Services specifically because both papers turn on the same structural point: a real, documented mechanism exists, and the missing piece is not the mechanism's existence but a governed record proving the mechanism did what it claims, in a specific instance, checkable by someone who was not the party making the claim.
This series remains open past its initial twelve-paper arc. This paper suggests a concrete next step for the architecture it describes rather than a new vertical: a governed compression record, computed with a real full-reference metric at the moment of encoding and anchored the way this series has described for every other signal, would close the gap this paper identifies directly, turning a repeated, unverified characterization into a checkable one.
VMAF, SSIM, PEAQ, and POLQA are the real, standardized, full-reference methods the video and audio industries use to verify a quality-preservation claim, and all of them require the original to check against. This series has used a version of MindAptiv's own quality claim four times, for cinema, live delivery, spatial audio, and now general web payload, without ever citing an independent score. This is Signal Paper XVIII.
Request Platform Access → Full White Paper Series