Normalize Filters! Classical Wisdom for Deep Vision Article Swipe
Classical image filters, such as those for averaging or differencing, are carefully normalized to ensure consistency, interpretability, and to avoid artifacts like intensity shifts, halos, or ringing. In contrast, convolutional filters learned end-to-end in deep networks lack such constraints. Although they may resemble wavelets and blob/edge detectors, they are not normalized in the same or any way. Consequently, when images undergo atmospheric transfer, their responses become distorted, leading to incorrect outcomes. We address this limitation by proposing filter normalization, followed by learnable scaling and shifting, akin to batch normalization. This simple yet effective modification ensures that the filters are atmosphere-equivariant, enabling co-domain symmetry. By integrating classical filtering principles into deep learning (applicable to both convolutional neural networks and convolution-dependent vision transformers), our method achieves significant improvements on artificial and natural intensity variation benchmarks. Our ResNet34 could even outperform CLIP by a large margin. Our analysis reveals that unnormalized filters degrade performance, whereas filter normalization regularizes learning, promotes diversity, and improves robustness and generalization.
Related Topics
Concepts
No concepts available.
Metadata
- Type
- preprint
- Language
- en
- Landing Page
- http://arxiv.org/abs/2506.04401
- https://arxiv.org/pdf/2506.04401
- OA Status
- green
- OpenAlex ID
- https://openalex.org/W4416131895
All OpenAlex metadata
Raw OpenAlex JSON
- OpenAlex ID
-
https://openalex.org/W4416131895Canonical identifier for this work in OpenAlex
- DOI
-
https://doi.org/10.48550/arxiv.2506.04401Digital Object Identifier
- Title
-
Normalize Filters! Classical Wisdom for Deep VisionWork title
- Type
-
preprintOpenAlex work type
- Language
-
enPrimary language
- Publication year
-
2025Year of publication
- Publication date
-
2025-06-04Full publication date if available
- Authors
-
Stella X. YuList of authors in order
- Landing page
-
https://arxiv.org/abs/2506.04401Publisher landing page
- PDF URL
-
https://arxiv.org/pdf/2506.04401Direct link to full text PDF
- Open access
-
YesWhether a free full text is available
- OA status
-
greenOpen access status per OpenAlex
- OA URL
-
https://arxiv.org/pdf/2506.04401Direct OA link when available
- Cited by
-
0Total citation count in OpenAlex
Full payload
| id | https://openalex.org/W4416131895 |
|---|---|
| doi | https://doi.org/10.48550/arxiv.2506.04401 |
| ids.doi | https://doi.org/10.48550/arxiv.2506.04401 |
| ids.openalex | https://openalex.org/W4416131895 |
| fwci | |
| type | preprint |
| title | Normalize Filters! Classical Wisdom for Deep Vision |
| biblio.issue | |
| biblio.volume | |
| biblio.last_page | |
| biblio.first_page | |
| is_xpac | False |
| apc_list | |
| apc_paid | |
| language | en |
| locations[0].id | pmh:oai:arXiv.org:2506.04401 |
| locations[0].is_oa | True |
| locations[0].source.id | https://openalex.org/S4306400194 |
| locations[0].source.issn | |
| locations[0].source.type | repository |
| locations[0].source.is_oa | True |
| locations[0].source.issn_l | |
| locations[0].source.is_core | False |
| locations[0].source.is_in_doaj | False |
| locations[0].source.display_name | arXiv (Cornell University) |
| locations[0].source.host_organization | https://openalex.org/I205783295 |
| locations[0].source.host_organization_name | Cornell University |
| locations[0].source.host_organization_lineage | https://openalex.org/I205783295 |
| locations[0].license | |
| locations[0].pdf_url | https://arxiv.org/pdf/2506.04401 |
| locations[0].version | submittedVersion |
| locations[0].raw_type | text |
| locations[0].license_id | |
| locations[0].is_accepted | False |
| locations[0].is_published | False |
| locations[0].raw_source_name | |
| locations[0].landing_page_url | http://arxiv.org/abs/2506.04401 |
| locations[1].id | doi:10.48550/arxiv.2506.04401 |
| locations[1].is_oa | True |
| locations[1].source.id | https://openalex.org/S4306400194 |
| locations[1].source.issn | |
| locations[1].source.type | repository |
| locations[1].source.is_oa | True |
| locations[1].source.issn_l | |
| locations[1].source.is_core | False |
| locations[1].source.is_in_doaj | False |
| locations[1].source.display_name | arXiv (Cornell University) |
| locations[1].source.host_organization | https://openalex.org/I205783295 |
| locations[1].source.host_organization_name | Cornell University |
| locations[1].source.host_organization_lineage | https://openalex.org/I205783295 |
| locations[1].license | |
| locations[1].pdf_url | |
| locations[1].version | |
| locations[1].raw_type | article |
| locations[1].license_id | |
| locations[1].is_accepted | False |
| locations[1].is_published | |
| locations[1].raw_source_name | |
| locations[1].landing_page_url | https://doi.org/10.48550/arxiv.2506.04401 |
| indexed_in | arxiv, datacite |
| authorships[0].author.id | https://openalex.org/A5042014034 |
| authorships[0].author.orcid | https://orcid.org/0000-0002-3507-5761 |
| authorships[0].author.display_name | Stella X. Yu |
| authorships[0].author_position | middle |
| authorships[0].raw_author_name | Yu, Stella X. |
| authorships[0].is_corresponding | True |
| has_content.pdf | False |
| has_content.grobid_xml | False |
| is_paratext | False |
| open_access.is_oa | True |
| open_access.oa_url | https://arxiv.org/pdf/2506.04401 |
| open_access.oa_status | green |
| open_access.any_repository_has_fulltext | False |
| created_date | 2025-10-10T00:00:00 |
| display_name | Normalize Filters! Classical Wisdom for Deep Vision |
| has_fulltext | False |
| is_retracted | False |
| updated_date | 2025-11-28T05:54:21.216338 |
| primary_topic | |
| cited_by_count | 0 |
| locations_count | 2 |
| best_oa_location.id | pmh:oai:arXiv.org:2506.04401 |
| best_oa_location.is_oa | True |
| best_oa_location.source.id | https://openalex.org/S4306400194 |
| best_oa_location.source.issn | |
| best_oa_location.source.type | repository |
| best_oa_location.source.is_oa | True |
| best_oa_location.source.issn_l | |
| best_oa_location.source.is_core | False |
| best_oa_location.source.is_in_doaj | False |
| best_oa_location.source.display_name | arXiv (Cornell University) |
| best_oa_location.source.host_organization | https://openalex.org/I205783295 |
| best_oa_location.source.host_organization_name | Cornell University |
| best_oa_location.source.host_organization_lineage | https://openalex.org/I205783295 |
| best_oa_location.license | |
| best_oa_location.pdf_url | https://arxiv.org/pdf/2506.04401 |
| best_oa_location.version | submittedVersion |
| best_oa_location.raw_type | text |
| best_oa_location.license_id | |
| best_oa_location.is_accepted | False |
| best_oa_location.is_published | False |
| best_oa_location.raw_source_name | |
| best_oa_location.landing_page_url | http://arxiv.org/abs/2506.04401 |
| primary_location.id | pmh:oai:arXiv.org:2506.04401 |
| primary_location.is_oa | True |
| primary_location.source.id | https://openalex.org/S4306400194 |
| primary_location.source.issn | |
| primary_location.source.type | repository |
| primary_location.source.is_oa | True |
| primary_location.source.issn_l | |
| primary_location.source.is_core | False |
| primary_location.source.is_in_doaj | False |
| primary_location.source.display_name | arXiv (Cornell University) |
| primary_location.source.host_organization | https://openalex.org/I205783295 |
| primary_location.source.host_organization_name | Cornell University |
| primary_location.source.host_organization_lineage | https://openalex.org/I205783295 |
| primary_location.license | |
| primary_location.pdf_url | https://arxiv.org/pdf/2506.04401 |
| primary_location.version | submittedVersion |
| primary_location.raw_type | text |
| primary_location.license_id | |
| primary_location.is_accepted | False |
| primary_location.is_published | False |
| primary_location.raw_source_name | |
| primary_location.landing_page_url | http://arxiv.org/abs/2506.04401 |
| publication_date | 2025-06-04 |
| publication_year | 2025 |
| referenced_works_count | 0 |
| abstract_inverted_index.a | 140 |
| abstract_inverted_index.By | 103 |
| abstract_inverted_index.In | 27 |
| abstract_inverted_index.We | 71 |
| abstract_inverted_index.as | 4 |
| abstract_inverted_index.by | 75, 80, 139 |
| abstract_inverted_index.in | 33, 51 |
| abstract_inverted_index.on | 126 |
| abstract_inverted_index.or | 8, 25, 54 |
| abstract_inverted_index.to | 13, 18, 68, 86, 112 |
| abstract_inverted_index.Our | 133, 143 |
| abstract_inverted_index.and | 17, 44, 83, 117, 128, 158, 161 |
| abstract_inverted_index.any | 55 |
| abstract_inverted_index.are | 10, 48, 98 |
| abstract_inverted_index.for | 6 |
| abstract_inverted_index.may | 41 |
| abstract_inverted_index.not | 49 |
| abstract_inverted_index.our | 121 |
| abstract_inverted_index.the | 52, 96 |
| abstract_inverted_index.yet | 91 |
| abstract_inverted_index.CLIP | 138 |
| abstract_inverted_index.This | 89 |
| abstract_inverted_index.akin | 85 |
| abstract_inverted_index.both | 113 |
| abstract_inverted_index.deep | 34, 109 |
| abstract_inverted_index.even | 136 |
| abstract_inverted_index.into | 108 |
| abstract_inverted_index.lack | 36 |
| abstract_inverted_index.like | 21 |
| abstract_inverted_index.same | 53 |
| abstract_inverted_index.such | 3, 37 |
| abstract_inverted_index.that | 95, 146 |
| abstract_inverted_index.they | 40, 47 |
| abstract_inverted_index.this | 73 |
| abstract_inverted_index.way. | 56 |
| abstract_inverted_index.when | 58 |
| abstract_inverted_index.avoid | 19 |
| abstract_inverted_index.batch | 87 |
| abstract_inverted_index.could | 135 |
| abstract_inverted_index.image | 1 |
| abstract_inverted_index.large | 141 |
| abstract_inverted_index.their | 63 |
| abstract_inverted_index.those | 5 |
| abstract_inverted_index.become | 65 |
| abstract_inverted_index.ensure | 14 |
| abstract_inverted_index.filter | 77, 152 |
| abstract_inverted_index.halos, | 24 |
| abstract_inverted_index.images | 59 |
| abstract_inverted_index.method | 122 |
| abstract_inverted_index.neural | 115 |
| abstract_inverted_index.simple | 90 |
| abstract_inverted_index.vision | 119 |
| abstract_inverted_index.address | 72 |
| abstract_inverted_index.degrade | 149 |
| abstract_inverted_index.ensures | 94 |
| abstract_inverted_index.filters | 30, 97, 148 |
| abstract_inverted_index.leading | 67 |
| abstract_inverted_index.learned | 31 |
| abstract_inverted_index.margin. | 142 |
| abstract_inverted_index.natural | 129 |
| abstract_inverted_index.reveals | 145 |
| abstract_inverted_index.scaling | 82 |
| abstract_inverted_index.shifts, | 23 |
| abstract_inverted_index.undergo | 60 |
| abstract_inverted_index.whereas | 151 |
| abstract_inverted_index.Although | 39 |
| abstract_inverted_index.ResNet34 | 134 |
| abstract_inverted_index.achieves | 123 |
| abstract_inverted_index.analysis | 144 |
| abstract_inverted_index.enabling | 100 |
| abstract_inverted_index.filters, | 2 |
| abstract_inverted_index.followed | 79 |
| abstract_inverted_index.improves | 159 |
| abstract_inverted_index.learning | 110 |
| abstract_inverted_index.networks | 35, 116 |
| abstract_inverted_index.promotes | 156 |
| abstract_inverted_index.resemble | 42 |
| abstract_inverted_index.ringing. | 26 |
| abstract_inverted_index.wavelets | 43 |
| abstract_inverted_index.Classical | 0 |
| abstract_inverted_index.artifacts | 20 |
| abstract_inverted_index.averaging | 7 |
| abstract_inverted_index.blob/edge | 45 |
| abstract_inverted_index.carefully | 11 |
| abstract_inverted_index.classical | 105 |
| abstract_inverted_index.co-domain | 101 |
| abstract_inverted_index.contrast, | 28 |
| abstract_inverted_index.effective | 92 |
| abstract_inverted_index.filtering | 106 |
| abstract_inverted_index.incorrect | 69 |
| abstract_inverted_index.intensity | 22, 130 |
| abstract_inverted_index.learnable | 81 |
| abstract_inverted_index.learning, | 155 |
| abstract_inverted_index.outcomes. | 70 |
| abstract_inverted_index.proposing | 76 |
| abstract_inverted_index.responses | 64 |
| abstract_inverted_index.shifting, | 84 |
| abstract_inverted_index.symmetry. | 102 |
| abstract_inverted_index.transfer, | 62 |
| abstract_inverted_index.variation | 131 |
| abstract_inverted_index.artificial | 127 |
| abstract_inverted_index.detectors, | 46 |
| abstract_inverted_index.distorted, | 66 |
| abstract_inverted_index.diversity, | 157 |
| abstract_inverted_index.end-to-end | 32 |
| abstract_inverted_index.limitation | 74 |
| abstract_inverted_index.normalized | 12, 50 |
| abstract_inverted_index.outperform | 137 |
| abstract_inverted_index.principles | 107 |
| abstract_inverted_index.robustness | 160 |
| abstract_inverted_index.(applicable | 111 |
| abstract_inverted_index.atmospheric | 61 |
| abstract_inverted_index.benchmarks. | 132 |
| abstract_inverted_index.integrating | 104 |
| abstract_inverted_index.regularizes | 154 |
| abstract_inverted_index.significant | 124 |
| abstract_inverted_index.consistency, | 15 |
| abstract_inverted_index.constraints. | 38 |
| abstract_inverted_index.improvements | 125 |
| abstract_inverted_index.modification | 93 |
| abstract_inverted_index.performance, | 150 |
| abstract_inverted_index.unnormalized | 147 |
| abstract_inverted_index.Consequently, | 57 |
| abstract_inverted_index.convolutional | 29, 114 |
| abstract_inverted_index.differencing, | 9 |
| abstract_inverted_index.normalization | 153 |
| abstract_inverted_index.normalization, | 78 |
| abstract_inverted_index.normalization. | 88 |
| abstract_inverted_index.transformers), | 120 |
| abstract_inverted_index.generalization. | 162 |
| abstract_inverted_index.interpretability, | 16 |
| abstract_inverted_index.convolution-dependent | 118 |
| abstract_inverted_index.atmosphere-equivariant, | 99 |
| cited_by_percentile_year | |
| corresponding_author_ids | https://openalex.org/A5042014034 |
| countries_distinct_count | 0 |
| institutions_distinct_count | 1 |
| citation_normalized_percentile |